PUBLISHED BY

Microsoft Press
A Division of Microsoft Corporation
One Microsoft Way
Redmond, Washington 98052-6399
Copyright © 2004 by David West
All rights reserved. No part of the contents of this book may be reproduced or transmitted in any form or by
any means without the written permission of the publisher.
Library of Congress Cataloging-in-Publication Data pending.

Printed and bound in the United States of America.

1 2 3 4 5 6 7 8 9 QWE 8 7 6 5 4 3
Distributed in Canada by H.B. Fenn and Company Ltd.
A CIP catalogue record for this book is available from the British Library.
Microsoft Press books are available through booksellers and distributors worldwide. For further information
about international editions, contact your local Microsoft Corporation office or contact Microsoft Press
International directly at fax (425) 936-7329. Visit our Web site at www.microsoft.com/mspress. Send comments
to mspinput@microsoft.com.
Microsoft, Visual Basic, Visual C++, Visual C#, Visual Studio, and Windows are either registered
trademarks or trademarks of Microsoft Corporation in the United States and/or other countries. Other
product and company names mentioned herein may be the trademarks of their respective owners.
The example companies, organizations, products, domain names, e-mail addresses, logos, people, places,
and events depicted herein are fictitious. No association with any real company, organization, product,
domain name, e-mail address, logo, person, place, or event is intended or should be inferred.
This book expresses the author’s views and opinions. The information contained in this book is provided
without any express, statutory, or implied warranties. Neither the authors, Microsoft Corporation, nor its
resellers or distributors will be held liable for any damages caused or alleged to be caused either directly
or indirectly by this book.
Acquisitions Editor: Robin Van Steenburgh and Linda Engelman
Project Editor: Denise Bankaitis and Devon Musgrave
Indexer: Shawn Peck

Body Part No. X10-25675

Table of Contents
Acknowledgments vii
Preface ix
Introduction xvii
1 Object Thinking 1
Observing the Object Difference 2
Object Thinking = Think Like an Object 12
Problem = Solution 16
Object Thinking and Agile Development Practices 18
Values 19
Selected Practices 22
Thinking Is Key 24
Software Development Is a Cultural Activity 25
Onward 30
2 Philosophical Context 33
Philosophy Made Manifest—Dueling Languages 36
SIMULA 38
C++ 41
Smalltalk 43
Formalism and Hermeneutics 48
Postmodern Critiques 58
3 From Philosophy to Culture 63
Four Presuppositions 66
One: Everything is an object. 66
Two: Simulation of a problem domain drives
object discovery and definition. 71
Three: Objects must be composable. 78
Four: Distributed cooperation and communication must replace
hierarchical centralized control as an organizational paradigm. 81
Object Principles—Software Principles 83
Cooperating Cultures 87

iii
iv Table of Contents

4 Metaphor: Bridge to the Unfamiliar 91

The Lego Brick Metaphor 96
The Object-as-Person Metaphor 101
Software as Theater; Programmers as Directors 108
Ants, Not Autocrats 112
Two Human-Derived Metaphors 113
Inheritance 114
Responsibility 115
Thinking Like an Object 116
5 Vocabulary: Words to Think With 117
Essential Terms 121
Object 121
Responsibility 123
Message 128
Interface (Protocol) 129
Extension Terms 130
Collaboration and Collaborator 130
Class 130
Class Hierarchy (Library) 132
Abstract/Concrete 133
Inheritance 133
Delegation 139
Polymorphism 140
Encapsulation 141
Component 142
Framework 142
Pattern 143
Implementation Terms 145
Method 145
Variable 145
Late/Dynamic Binding 146
 Table of Contents v

Auxiliary Concepts 147

Domain 147
Business Requirement 149
Business Process Reengineering 149
Application 149
6 Method, Process, and Models 151
Two Decades of Object Methodology 153
Purpose and Use of Method 159
A Syncretic Approach 164
Models 168
Semantic Net 169
Object Cubes 173
Interaction Diagram 175
Static Relation Diagram 178
Object State Chart 181
7 Discovery 183
Domain Understanding 185
Domain Anthropology 186
Object Definition 200
Heuristics 212
8 Thinking Toward Design 219
Object Internals 220
Knowledge Required 221
Message Protocol 227
Message Contracts 234
State Change Notification 236
Object Appearance 240
Occasions Requiring an Appearance 241
Object State, Object Constraints 245
vi Table of Contents

9 All the World’s a Stage 247

Static Relationships 251
Is-a-Kind-of Relationship 251
Collaborates-with Relationship 254
Situational Relationship 256
Dynamic Relationships 273
Scripts 274
Event Dispatching 277
Constraints 281
Self-Evaluating Rules 282
Implementation 286
Methods 287
Knowledge Maintenance Objects 288
Development at the Speed of Thought 291
10 Wrapping Up 293
Vexations 294
The Impedance Mismatch Problem 294
A Problem with GUIs 297
Extensions 299
Frameworks 299
Object-Based Evocative Architecture 302
Provocation—The Future of Objects 305

Bibliography 309

Index 321
Acknowledgments
One name appears on the cover as author of this book, hiding the fact that
every book is a collaborative effort involving scores of contributors. Although it
is impossible to acknowledge and thank everyone who played a role, I can and
must name some individuals. Each is owed my personal thanks, and each
stands as a proxy for many others whom I acknowledge in my mind and heart.
Mary—without whom this would never have been done. My muse, my
friend, my spouse of twenty-one years.
Maurine, Bob, Sara, Ryan, and Kathleen—my family, whose support was
required and freely given.
Kevin—the best programmer with whom I have ever worked. You
proved, sometimes after a lot of discussion as to why it was impossible, that my
crazy ideas could be implemented.
Tom, Pam, Don, Dion, Julie, Kyle, Dave, Steve, and J.P.—our initial con-
tact was as student and professor, but you became colleagues and friends
and represent the hundreds of St. Thomas alumni who helped shape object
thinking and then applied it in the real world with notable success.
Tom and Ken—your technical review, insightful comments, correction of
errors, and honest advice were invaluable. This would not have been as useful
a book without your assistance.
Linda, Devon, Denise, Robin, Shawn, Joel, Sandi, and Elizabeth—the
editors and staff at Microsoft Press who smoothly dealt with all the technical
aspects of getting a book out the door (including an author with deadline
issues). An author could not have found a friendlier, more helpful, or more
professional group of editors and craftspeople with whom to work.

vii
Preface

A Different (and Possibly Controversial) Kind of Software Book

This book will be deliberately different from almost any other object analysis/
design, component-based development, software development methodology,
or extreme programming (XP) book you may have encountered. It’s also likely
to be controversial, which is not intended but is, perhaps, inevitable. Several
factors contribute to this book’s differences:

■ The reader will be asked to read and digest a lot of history and
philosophy before embarking on the more pragmatic aspects of
object thinking.
■ The author will unabashedly, adamantly, and consistently advocate
behavior as the key concept for discovering, describing, and design-
ing objects and components.
■ The centrality of CRC (Class-Responsibility-Collaborator) cards as an
object thinking device or tool will likely be viewed as anachronistic
and irrelevant since UML (Unified Modeling Language) has achieved
the status of de facto standard.
■ The apparent indifference, or even antagonism, toward formalism
and the formal approaches to software specification that are
intrinsic to the behavioral approach will concern some readers,
especially those trained in computer science departments at
research universities.
■ The emphasis on analysis and conceptualization—that is, on think-
ing—instead of on implementation detail might strike some readers
as academic.

It will take the rest of the book to explain why these differences are
important and necessary, but the motivation behind them is found in this 1991
quote from Grady Booch:

ix
x Object Thinking

Let there be no doubt that object-oriented design is

fundamentally different than traditional structured design
approaches: it requires different ways of thinking about
decomposition, and it produces software architectures that are
largely outside the realm of the structured design culture.
“Different ways of thinking” is the key phrase here, and the word culture
is equally important. The history of objects in software development is charac-
terized by the mistaken notion that the object difference was to be found in a
computer language or a software module specification. But objects are funda-
mentally different because they reflect different ideas—particularly about
decomposition—and because they reflect a different set of values and a differ-
ent worldview (that is, culture) from that of traditional software development.
Understanding objects requires understanding the philosophical roots and his-
torical milestones from which objects emerged.
Readers will be expected to review and reflect on two competing philo-
sophical traditions—formalism and hermeneutics-postmodernism—that are
responsible for the value systems and ideas leading to the competition and
heated disagreements between advocates of software engineering (who value
formalism) and advocates of XP/agile methodologies/behavioral objects (who
value hermeneutics). A quick look at the history of programming in general will
show that XP and agile methodologies are but the latest manifestations in this
long-standing philosophical feud, which has also influenced our (mis)under-
standing of objects and patterns.
Distractions to object thinking, such as which programming language is
“best,” will be shown to be mostly irrelevant with a brief recap of classical
languages claiming to be object-oriented. This historical excursion will also
show you how ideas become manifest in tools—in this case programming
languages—and how philosophical principles and cultural values shape soft-
ware development methods and processes.
Metaphor and vocabulary play a major role in shaping object thinking.
Metaphors are essential for bridging the familiar and the unfamiliar, and selec-
tion of the “right” metaphors is critical for further development of fundamental
ideas. Likewise, vocabulary provides us with one of the essential tools for think-
ing and communicating. It’s important to understand why object advocates
elected to use a different vocabulary for object thinking and why it’s inappro-
priate to project old vocabulary onto new concepts.
As you might expect by now, the prominent role of formally defined mod-
els, methods, and processes will be discounted in this book; their claim to be
repositories of objective truth will be challenged. This book will suggest that
methods are little more than helpful reminders of “things to think about” and
that models are only a form of external short-term memory useful in the context
 Preface xi

of a particular group of people at a particular point in time. These biases will be

linked to XP/agile approaches.
Note that I’ll also provide applications of the ideas this book offers. I’ll
provide several examples as illustrations of how the ideas exhibit themselves
in practice.
However focused on a particular topic each section of this book might be,
an overarching bias or perspective colors every discussion: behaviorism. If a
single word could be used to capture the “different ways of thinking about
decomposition” noted by Booch, it would be behavior. The centrality of behav-
ior will be most evident when we consider and compare traditional definitions
and models of objects.
A majority of the published books on objects mention using behavior as
the criterion for conceptualizing and defining objects. Many claim to present
behavioral approaches to object development. However, with one or two
exceptions,1 they provide object definitions and specification models that owe
more to traditional concepts of data and function than to behavior. Even when
behavior is used to discover candidate objects, that criterion is rapidly set aside
in favor of detailed discussions of data attributes, member functions, and class/
entity relationships. This book’s focus on behavior, on its aspects and implica-
tions, is consistent, I believe, with both the origins of object ideas and the
design tradition assumed by XP/agile.
Material in this book is presented in a matter-of-fact style, as if everything
stated were unequivocally correct. I’ll cite alternative ideas and approaches, but
I’ll make little effort to incorporate those alternatives or to discuss them in
detail. This is not because I am unaware of other viewpoints or because I am
dismissive of those viewpoints. Alternative ideas are best expressed by their
advocates. To the extent that this book is used as a classroom text, discussion
of alternative viewpoints and alternative texts is the purview of the instructor.

Paths and Destinations

The destination is easy—to become a better software developer. Or maybe not
so easy since “better” requires more than the addition of yet another skill or
technique to one’s repertoire. Techniques, tools, skills, and facts are part of
becoming better but are insufficient in themselves. Integration with existing
skills, the transformation of facts into knowledge (even wisdom), and the ability
to base one’s future actions on principles and ideas instead of rote procedures
are all essential to becoming better.

1. Wirfs-Brock, Weiner, and Wilkerson’s Objected Oriented Design offers a behavior-based method, as
does Wilkinson’s Using CRC Cards.
xii Object Thinking

This book is intended to offer one path to becoming a better developer.

Specifically, a better agile developer; and more specifically yet, a better extreme
programmer. By better, I mean more like the acknowledged masters of agile
development and extreme programming: Ward Cunningham, Kent Beck, Ron
Jeffries, Alistair Cockburn, Bob Martin, Ken Auer … (Please fill in the rest of the
top 20 to reflect your own heroes.)
These people are not masters merely because they practice certain tech-
niques or exhibit certain skills. They embody, they live, they exude shared val-
ues and principles. They share similar experiences and have learned common
lessons from those experiences. They share a worldview. Given a common
anthropological definition of culture—“shared, socially learned knowledge and
patterns of behavior”2—it is reasonable to assert that the agile and extreme pro-
gramming masters constitute a culture—a subculture, actually—of software
developers. If one aspires to be like them, one must become a member of that
culture.
The process of learning a culture—enculturation—is partly explicit but
mostly implicit. The explicit part can be put into books and taught in seminars
or classrooms. Most of culture is acquired by a process of absorption—by living
and practicing the culture with those who already share it. No book, including
this one, can replace the need to live a culture. But it is possible to use a book
as a means of “sensitizing,” of preparing, a person for enculturation—shorten-
ing the time required to understand and begin integrating lived experiences.
That is the modest goal of this author for this book.

Who Should Read This Book

A conscious effort has been made to make this book useful to a wide audience
that includes professional developers and postsecondary students as well as
anyone with an interest in understanding object-oriented thinking. It’s assumed
that the average reader will have some degree of familiarity with software
development, either coursework or experience. It’s further assumed that the
reader will have been taught or will have experience using the vocabulary,
models, and methods prevalent in mainstream software development. This
book focuses on presenting objects and XP and not on providing background
and details of mainstream development.

2. Peoples, James, and Garrick Bailey. Humanity: an Introduction to Cultural Anthropology. Belmont,
CA: Wadsworth/Thompson Learning, 2000.
 Preface xiii

The organization of the book, front-loading the philosophical and histori-

cal material, might present some problems for the most pragmatically oriented
reader, the one looking for technique and heuristics to apply immediately. The
introduction attempts to make a case for reading the book in the order pre-
sented, but it’s OK to skip ahead and come back to the early chapters to under-
stand the ideas behind the technique.
Three caveats help to further define the audience for this book:

■ This is not a programming text. Some limited examples of pseudocode

(usually having the flavor of Smalltalk, the language most familiar to
the author) are presented when they can illuminate a concept or
principle of development.
■ The book, however, is expressly intended for programmers, espe-
cially those using Java and C++. It’s hoped that this book will facili-
tate their work, not by providing tricks and compiler insights but by
providing conceptual foundations. Languages such as C++ and Java
require significant programmer discipline if they are to be used to
create true objects and object applications. That discipline must, in
turn, be grounded in the kind of thinking presented in this book.
■ Although this book stresses the philosophy and history of objects/
components, the author’s overarching goal is assisting people in the
development of pragmatic skills.

Note If this text is used in an academic course (undergraduate or

graduate), roughly 40 percent of the available class time should be
devoted to workshop activities. Software development of any kind is
learned through experience, but objects, because they are new and
different, require even greater amounts of reflection and practice.
Another characteristic of an ideal course is interaction and discussion
of multiple viewpoints. This book is best used in conjunction with other
method books (particularly any of the excellent books on UML or RUP)
that present alternative viewpoints and, of course, with at least one of
the XP or agile texts, which will add depth to what is presented here.

Portions of the historical and philosophical material in this book will be

familiar to those readers with their own extensive personal experience as prac-
titioners. To them, the material might appear to be rehashed old arguments. But
xiv Object Thinking

most readers will not share either that experience or that awareness, and they
need to know, explicitly, how those ideas have shaped the profession as we
see it today. I hope that even the most experienced practitioners will see some
new insights or have old insights brought to the forefront of their minds when
reading this material.
One final note. Readers of this book will tend to be professional software
developers or those aspiring to become professionals. Both groups share a
character trait—an eagerness to build, to practice. They are ready to “just do it.”
To them is directed the final message of this introductory chapter: please
be patient.
As a profession, we also tend to be abysmally ignorant of our own history.
In the words of the oft quoted, misquoted, and paraphrased philosopher,
George Santayana, “Those who cannot remember the past are condemned to
repeat it.” In computing, thanks to Moore’s Law, our repetition is disguised a bit
by scale—mainframe mistakes replicated on desktops, then notebooks, then
PDAs and phones, and then molecular computers.
As a profession, we tend to ignore the various influences—such as cul-
ture, philosophy, psychology, economics, and sheer chance—that have shaped
the practice of software development.
It is my belief that we cannot improve our intrinsic abilities as object and
agile software developers without an examination, however brief, of the histor-
ical and philosophical presuppositions behind our profession. We should not,
and cannot if we want them to truly understand, bring new members into our
profession without providing them with more than tools, methods, and pro-
cesses that they use by rote.

How This Book Is Organized

Readers are encouraged to proceed through the book in sequence, to accept
that each chapter lays a foundation for the one that follows. Overlaying the
order are several implicit logical divisions, including the following:

■ The introduction and Chapter 1, “Object Thinking,” advance argu-

ments for why understanding the background and history of ideas is
a necessary step in the successful application of those ideas. They
also argue that both objects and extreme programming (agile methods)
share common foundations.
■ Chapter 2, “Philosophical Context,” and Chapter 3, “From Philosophy
to Culture,” provide a foundational context, partly based in philoso-
phy and partly in history.
 Preface xv

■ Chapter 4, “Metaphor: Bridge to the Unfamiliar,” introduces key ideas

and meta-ideas (metaphors as ideas about how to explain ideas, sim-
ply put).
■ Chapter 5, “Vocabulary: Words to Think With,” introduces vocabulary
and explains why object thinking requires a different vocabulary for
things that appear to be familiar but that use old labels.
■ Chapter 6, “Method, Process, and Models,” revisits commonalities
between object and agile ideas and the relationship of those ideas to
the notion of method and process.
■ Chapter 7, “Discovery,” Chapter 8, “Thinking Toward Design,” and
Chapter 9, “All the World’s a Stage,” apply the ideas of previous
chapters and provide examples of object thinking in action.
■ Chapter 10, “Wrapping Up,” adds a coda with short explorations of
how object thinking can be extended and coordinated with unavoid-
able nonobject worlds, and what the ultimate outcome of object
thinking might be.

I encourage you to engage the book in the order presented, but feel free
to skip Chapter 1 if you’re prepared to take on faith the assertions made about
objects and object thinking in subsequent chapters.
Many readers will be a bit anxious to “get to the good stuff,” to see how
the book’s ideas are applied or how they manifest themselves in practice
(which is shown in Chapters 7 through 9). While I firmly believe that the
book’s initial chapters—as far afield as they might appear to be—need to be
covered first, I also recognize the need to at least foreshadow application. To
this end, I offer a continuing sidebar—the “Forward Thinking” sidebar—in
these initial chapters. This sidebar presents a sample application in bits and
pieces as a means of illustrating important ideas while satisfying the craving
for pragmatic examples.
Introduction

The time: 1968. A software crisis has been declared. Part of the crisis derives
from the fact that more software is needed than there are developers to produce
it. The other part of the crisis is the abysmal record of development efforts.
More than half of the projects initiated are canceled, and less than 20 percent of
projects are successfully completed, meaning that the time and cost overruns
associated with those projects were less than 100 percent and the software was
actually used to support the business.
The time: 2003. We have a lot of developers—part of the software crisis
has passed for the moment. Many of them are located in other countries
because too many managers seem to believe that developers are developers
and therefore go for greater numbers at lesser cost.
The skills, abilities, attitudes, and aptitudes of the development commu-
nity are, unfortunately, suspect. The “development community” means the
entire development community—from programmers to senior managers—in
the United States and the rest of the world. It is still the case that almost half of
all projects initiated are not completed. Those that are completed almost always
incur significant cost overruns. Quality—the lack thereof—is still a major issue.
Bloated, inefficient, bug-ridden, user-unfriendly, and marginally useful—these
are still common adjectives used to describe software.
The lack of significant improvement is not for want of trying. Numerous
proposals have been advanced to improve software development.
Software engineering provides an umbrella label for a spectrum of
improvement efforts ranging from structured programming to formal (mathe-
matical and logical) foundations for program specification to UML (unified
modeling language) to ISO 9000 certification and CMM (capability maturity
model) formalizations of the development process.
Sharing the goal of software improvement but rejecting the assumptions
implicit in software engineering, several alternative development approaches
have been championed. Creativity and art have been suggested as better meta-
phors than engineering. Software craftsmanship1 is the current incarnation of
this effort. Iterative development has as long a history as the linear-phased
development at the heart of software engineering.

1. McBreen, Peter. Software Craftsmanship: The New Imperative. Addison-Wesley, 2001.

xvii
xviii Object Thinking

Extreme programming (XP) and agile software development represent a

contemporary attempt to redefine and improve software and the practice by
which that software comes into existence. XP/agile represents an alternative to
software engineering. Further, XP/agile challenges the assumptions, the core
values, and the worldview upon which software engineering is predicated. It is
not surprising, therefore, that those with a vested interest in software engineer-
ing are trying very hard to marginalize (“it’s OK for small projects by small
teams that are not engaged in mission-critical projects”), co-opt (“XP is just a
subset of RUP2—with a different vocabulary”), or dismiss (“XP is just a tantrum
staged by a few petulant out-of-work Smalltalk programmers angry at the
demise of their favorite programming language”).
This book is motivated by a set of curious questions; it is based on the
belief that “better people” are absolutely essential and that they can be nur-
tured; it is grounded in the conviction that object-oriented ideas and principles
are poorly understood; and it is premised on the belief that XP (and other agile
methods) and object thinking are inextricably entwined—each requires the
other if they are to be successful.

Curiosities
Consider the following curious questions: Why is it that iterative development
has been acknowledged—even by those proposing sequential (waterfall)
development—as the “right” way to produce software, and yet

■ XP is seen as a “novel” and even “radical” approach?

■ So few, except those who have attempted to install an agile
approach the past two or three years, “officially” use iterative
development?
■ There is so much resistance—by managers and professional
developers—to doing things the “right” (iterative) way?

2. RUP—Rational Unified Process, an attempt to standardize and formally define the software develop-
ment process. Created and advocated by the same individuals behind UML (unified modeling lan-
guage), most notably Grady Booch, Ivar Jacobson, and James Rumbaugh.
 Introduction xix

■ Discussions between proponents and opponents of agile approaches

are so heated and frequently emotional when both sides seem to
agree, on the surface, on so many things?
■ “Better people” has been recognized as the most promising silver
bullet for addressing the software crisis, and yet almost all of our
energy has been spent on creating better tools, methods, and pro-
cesses instead of better people?
■ Every effort to advance nonformal, iterative, artistic, and humane
ways to develop software seem to be resisted and then co-opted and
debased by formal software engineering?
■ So few developers seem to be able to adopt and practice innovations
such as objects and agile methods in a manner consistent with the
intent and example of those who first advanced the innovation?
(Don’t believe this is a fair statement? Then why, two to four years
after XP and agile were introduced, is that community spending so
much time and effort wrestling with questions of certification? Why
does Alan Kay say the object revolution has yet to occur?)

The long answer to these and similar questions is this book. The short
answer, and hopefully part of your motivation for reading this book, is that soft-
ware developers tend to be so focused on what and how that they forget to
explore questions of why.

The “People Issue”

Fact 1: The most important factor in software work is not the
tools and techniques used by the programmers but rather the
quality of the programmers themselves.

Fact 2: The best programmers are up to 28 times better than the

worst programmers, according to “individual differences”
research. Given that their pay is never commensurate, they are
the biggest bargains in the software field.
—Robert L. Glass3

3. Glass, Robert L., Facts and Fallacies of Software Engineering. Boston: Addison-Wesley, 2003.
xx Object Thinking

In his discussion of the foregoing facts, Glass notes that we have been
aware of these human differences since 1968:4

Nearly everyone agrees, at a superficial level, that people trump

tools, techniques, and process. And yet we keep behaving as if it
were not true. Perhaps it’s because people are a harder problem
to address than tools, techniques, and process.

We in the software field, all of us technologists at heart, would

prefer to invent new technologies to make our jobs easier. Even
if we know, deep down inside, that the people issue is a more
important one to work.
—Glass, 2003

This book is an attempt to address the “people issue.” Specifically, it is an

attempt to help developers improve their intrinsic abilities.

The Need for Better Developers

For decades, the profession of software development has relied on an implicit
promise: “Better tools, methods, and processes will allow you to create superior
software using average human resources.” Managers have been sold the idea
that their jobs will be made easier and less risky by adoption of strict formal
methods and processes—a kind of intellectual Taylorism. (Taylor, Gilbreth, and
others advanced the concept of scientific management, time and motion stud-
ies, and the modern assembly line. Their distrust of human workers led them to
the idea that imposed process and method could compensate for weaknesses
they felt to be innate in workers. The same attitude and philosophy are
assumed in the field of software engineering.) For the most part, these efforts
have failed miserably.
Instances of success exist, of course.5 You can find cases in which project
A using method X did, in fact, achieve notable success, but industry statistics as
a whole have failed to improve much since 1968, when software engineering
and scientific management were introduced as means for resolving the software
crisis. Abandoned projects, cost/time overruns, and bloated, buggy software
still dominate the landscape.

4. Sackman, H., W.I. Erikson, and E.E. Grant. “Exploratory Experimental Studies Comparing Online and
Offline Programming Performances.” Communications of the ACM, January 1968.
5. Some advocates of software engineering and process improvement will make claims of major suc-
cesses (SEI Report CMU/SEI-2001-SR-014).
 Introduction xxi

Even the most ardent advocates of software engineering (for example,

Dykstra, Boehm, and Parnas), as well as those most responsible for populariz-
ing software engineering among the corporate masses (Yourdon and Martin),
recognized the limits of formal approaches. In “A Rational Design Process: How
and Why to Fake It,”6 Parnas acknowledged the fact that highly skilled devel-
opers did not “do” development the way that so-called rational methods sug-
gested they should. Martin suggested that the best way to obtain high-quality
software—software that met real business needs on time and within budget—
was the use of special “SWAT” teams comprising highly skilled and greatly
rewarded developers doing their work with minimal managerial intervention.
The only consistently reliable approach to software development is, sim-
ply, good people. So why have so much attention and effort been devoted to
process and method? There have been at least three contributing reasons:

■ A widespread belief, partially justified, that not enough good people

were available to accomplish the amount of development that
needed to be done.
■ An unspoken but just as widely held belief that really good develop-
ers were not to be trusted—they could not be “managed,” they all
seemed to be “flaky” to some degree, and they did not exhibit the
loyalty to company and paycheck of “normal” employees. Really
“good” developers tended to be “artists,” and art was (is) not a good
word in the context of software development.
■ A suspicion, probably justified, that we did not (do not) know how
to “upgrade” average developers to superior developers except by
giving them lots of experience and hoping for the best.

It is no longer possible to believe that either method or process (or both

together) is an adequate substitute for better people. There is a resurgence of
interest—spurred most recently by XP and the core practices in other agile
methods—in how to improve the human developer. This interest takes many
forms, ranging from XP itself to redefinition of software development as a craft
(McBreen7) and a calling (West8, Denning9) instead of a career to software
apprenticeship (Auer10). Why do we need better developers? Because increasing

6. Parnas, David Lorge, and Paul C. Clements. “A Rational Design Process: How and Why to Fake It.”
IEEE Transactions on Software Engineering. Vol. SE-12, No. 2, February 1986.
7. McBreen, Peter. Software Craftsmanship: The New Imperative. Boston: Addison-Wesley, 2001.
8. West, David. “Enculturating Extreme Programmers” and “Educating Xgilistas.”
9. Denning, Peter. ACM editorial.
10.Ken Auer, www.rolemodelsoft.com.
xxii Object Thinking

the supply of highly skilled people—rather than only adhering to particular

methods and processes—is the only way to resolve the software crisis.

Note Extreme programming and agile methods derive from the

actual practice of software development rather than academic theory.
The critical difference between XP/agile and traditional methods is
their focus on attitudes, behaviors, culture, and adaptive heuristics
instead of formal technique and theory. If the “method” label is to be
attached to XP/agile, it should be in terms of a method for producing
better developers rather than a method for producing better software.
Better software comes from, and only from, better people.

Producing Better Developers

What makes a better developer? The traditional answer is experience. Most text-
books on software engineering and formal process contain a caveat to the
effect that extensive real-world experience is required before the tools and
techniques provided by method/process can be effectively utilized. In this con-
text, experience is just a code word for those aspects of development—philos-
ophy, attitude, practices, mistakes, and even emotions—that cannot be reduced
to syntactic representation and cookbook formulation in a textbook.
Today’s practitioners, and some theorists, are intensely interested in teas-
ing apart and understanding the complex elements behind the label “experi-
ence.” There is great desire to understand what developers actually “do” as
opposed to some a priori theory of what is appropriate for them to do. Models
are seen as communication devices, unique to a community of human commu-
nicators, instead of “vessels of objective truth” with a value that transcends the
context of their immediate use. Development is seen as a human activity—
hence a fallible activity—that must accommodate and ameliorate imperfection
and mistakes rather than seek to eliminate them. Communalism is valued over
individualism. Systems are seen as complex instead of just complicated, neces-
sitating a different approach and different insights than were required when
software developers merely produced artifacts that met specification.
All of this intense interest is producing results. Although the total picture
is still emerging, some facets of “the better developer” are coming into focus.
XP provides a foundation by identifying a set of discrete practices that can
actually be practiced. Any developer can actually do these practices and, simply
by doing them, become a better practitioner. They offer no grand theory, nor
 Introduction xxiii

are they derived from any theory. The justification for the XP approach is based
on two simple empirical observations: “We have seen master developers do
these things” and “We have seen less proficient developers do these things and
become better.” Although XP does make claims to improve both product and
process, these are side effects—one is tempted to say mere side effects—of the
improvement in the human developers.
XP/agile approaches provide a solid foundation, but just a foundation.
What “Xgilistas”11 do is but part of what makes them master developers. What
they think and how they think are critically important as well. XP relies on max-
imized communication and storytelling as a means for enculturating new devel-
opers in appropriate ways of thinking. Thinking includes a value system, a
history, a worldview, a set of ideas, and a context. And XP encompasses an oral
tradition that has not, as yet, been reduced to ink and paper (and maybe cannot
be so reduced). Aspiring Xgilistas must become conversant with all of this
before they can attain true “master” status.
One component of the oral tradition, of the history, and of the common
worldview is the use of object-oriented approaches to design and program-
ming. Unfortunately, this is seldom made explicit in the XP literature. The terms
object and object-oriented do not appear in any of the first five books in the
Addison-Wesley XP series—except once, and that occasion points to an incor-
rect page in the text. However, object vocabulary and concepts are abundantly
evident. This discrepancy merely confirms that object thinking is presupposed
by those advocating XP. The primary goal of this book is to provide one small
contribution to help those following the Xgilista path—specifically, a contribu-
tion in the area of object thinking.

Object Thinking
Thirty plus years have passed since Alan Kay coined the term object-oriented.
Almost all contemporary software developers describe their work using object
vocabulary and use languages and specification tools that lay claim to the
object label. The ubiquity of object terminology does not mean, however, that
everyone has mastered object thinking. Nor does the popularity of Java. Nor
does the de facto standardization of object modeling embodied in UML. A pre-
diction made by T. Rentsch (cited by Grady Booch in 199112) remains an accu-
rate description of today’s development and developers:

11. A neologism—useful for labeling those that embody one or more of the approaches that fall under
the “agile” label.
12. Booch, Grady. Object-Oriented Analysis and Design with Applications, Second Edition. Boston:
Addison-Wesley, 1993.
xxiv Object Thinking

My guess is that object-oriented programming will be in the

1980s what structured programming was in the 1970s.
Everyone will be in favor of it. Every manufacturer will promote
his products as supporting it. Every manager will pay lip service
to it. Every programmer will practice it (differently). And no
one will know just what it is.
In fact, the situation may be worse than Rentsch predicted. An argument
can be made that the contemporary mainstream understanding of objects is but
a pale shadow of the original idea. Further, it can be argued that the mainstream
understanding of objects is, in practice, antithetical to the original intent.
Clearly the behavioral approach to understanding objects has all but dis-
appeared. (Yes, UML does allow for a “responsibility” segment in its class dia-
gram and description, but this hardly offsets the dominant trend in UML to treat
objects as if they were “animated data entities” or “miniature COBOL pro-
grams.”) This fact is important because two of the leading advocates behind XP
were also the leading advocates of “behavioral objects.” Kent Beck and Ward
Cunningham invented the CRC card approach to finding and defining objects—the
most popular of the “behavioral methods.” Others deeply involved in the Agile
Alliance were also identified with “object behavioralism” and with Smalltalk, the
programming language that came closest to embodying behavioral objects.
It isn’t unreasonable to assume that the behavioral approach to under-
standing objects dominates the “object thinking” of many of the best XP practi-
tioners, in part because it was almost necessarily part of the oral tradition
passed on by Kent Beck as he initiated others into the XP culture.

Note Certain aspects of XP practice are simply incarnations of earlier

practices. XP “stories,” for example, which deal with a system function
actualized by a group of objects, are identical to the scenarios talked
about in conjunction with CRC cards. Stories that are factored to their
simplest form are nothing more than responsibilities. The only real dif-
ference between stories and scenarios/behaviors is the insistence that
stories are always, and only, told by the domain expert (the customer)
and always depict interactions in the problem domain. This was also
true of good scenarios, but in the object literature scenarios tended to
reflect implementation more than they did problem description.
 Introduction xxv

It’s also reasonable to assume that behavioral object thinking is only implicit
in the XP/agile culture because so few books or texts were ever written in sup-
port of this approach. Neither Beck nor Cunningham ever wrote such a book.
Rebecca Wirfs-Brock didn’t update her 1991 book describing behavior-based
object design until this year. Other efforts in this area (for example, Nancy Wilk-
erson’s book on CRC cards) were not widely known and have not been updated.
This is particularly unfortunate because the CRC card “method” as described
in the early 1990s did not incorporate all aspects of object thinking. In fact, I
believe that object thinking transcends the notion of method just as it transcended
programming languages. (You can do good object programming in almost any
language, even though some languages offer you more support than others.)
Object thinking requires more than an understanding of CRC cards as pre-
sented circa 1990. It also requires understanding some of the history and some
of the philosophical presuppositions behind object behavioralism, CRC cards,
and languages such as Smalltalk. It requires an understanding of the metaphors
that assist in good object thinking and an extension of the CRC card metaphor,
in particular, to include more than object identification and responsibility
assignment.
It is my hope that this book will promote such an understanding by cap-
turing at least part of the oral tradition of behavioral objects and making it
explicit.

XP and Object Thinking

This book is based on the following beliefs:

■ Agility, especially in the form of extreme programming, is essential if

the software development profession, and industry, are to improve
themselves.
■ XP offers a way to create “better developers” to improve the intrinsic
abilities of human beings to develop software for use by other
human beings—a way to produce “master” developers.
■ XP cannot be understood, and those practicing XP will not realize
the full potential of the approach, until they understand object think-
ing and the shared historical and philosophical roots of both object
thinking and XP core values and practices.
■ In particular, programmers and technical developers will fail to real-
ize the full potential of XP-based development without a thorough
understanding of object orientation—the “object thinking” promised
by the title of this book.
Object Thinking
Let there be no doubt that object-oriented design is fundamentally different
than traditional structured design approaches: it requires a different way of
thinking about decomposition, and it produces software architectures that are
largely outside the realm of the structured design culture.1
——Grady Booch

I have three main goals for this initial chapter:

■ To begin showing the reader why object thinking is different and

important
■ To introduce arguments as to why object thinking is particularly
valuable to developers hoping to master extreme programming or
other agile approaches2
■ To explain the emphasis on thinking instead of tools or technique; to
explain the relevance and importance of history, philosophy, and
metaphor—topics that are seldom the focus of books purporting to
improve your software development skills

1. Booch, Grady. Object-oriented Design: with Applications. Redwood City, CA: Benjamin/Cummings.
1991. ISBN: 0-8053-0091-0.
2. In the remainder of this book, I will generally use XP instead of XP and other agile methods as a short-
hand label for aspects common to all agile methods.

1
2 Object Thinking

The short quote from Grady Booch that opens this chapter contains three
key phrases that provide a framework for the discussion in the remainder of
this chapter:

■ “fundamentally different” Which raises the question of how

object-oriented design’s difference manifests itself and whether, in
some aspects of development, this difference is more obvious or
important than in other aspects.
■ “different way of thinking” Not only about decomposition but
about all aspects of development. Of course, this is the theme of this
entire book.
■ “structured design culture” By explicitly mentioning a structured
design culture, Grady implies the existence of a corresponding object
culture. There is an object culture and an extreme programming culture,
and both cultures share a common heritage. Culture is a critical factor in
how and what we think as human beings and as software developers.

Observing the Object Difference

Is Figure 1-1 or Figure 1-2 a better visual depiction of an object?

Encapsulation
Barrier

operationX

Data
operationY Structure

operationZ

Figure 1-1 A graphical object model using UML graphical syntax.

F01MQ01
 Chapter 1 Object Thinking 3

Figure 1-2 A person as visual metaphor of an object.

F01MQ02

A traditional developer will almost always pick the diagram in Figure 1-1.
(Ken Auer calls this a “soccer ball” diagram because the segmentation resem-
bles that of a soccer ball.) An object thinker, on the other hand, will pick the
photograph. The soccer ball diagram has the advantage of familiarity to a tradi-
tional developer. It employs terms that are readily understood, and more impor-
tant, it reinforces premises critical to the mindset of traditional developers and
computer scientists. The diagram embodies Dykstra’s definition of a program as
data structures plus algorithms along with the idea of black box (enabled by the
encapsulation barrier) modularization.
Anthropomorphization—a big word for the practice of projecting human
characteristics onto objects—is fundamental to object thinking, which accounts
for the selection of the person depicted in the photograph by object thinkers.
As a thinking device—a metaphor or mental model—the soccer ball diagram
impedes object thinking by perpetuating old mental habits. For example, if the
diagram is an accurate depiction of an object, what is the difference between an
object and a COBOL program? There is none. A COBOL program encapsulates
data (data division) and operations (procedure division) and allows communi-
cation among programs (communication division). Object development—using
this model—will have a tough time being anything more than the creation of
lots of tiny COBOL programs.
4 Object Thinking

Note A COBOL program is, of course, a very large object. Inside the
program is a lot of messy stuff that’s not at all object-like. Object
COBOL differs from COBOL primarily in its ability to create smaller
programs—closer to the scale of classes/objects in other program-
ming languages—and enhancement of the communication division.
No substantial changes have been made—object COBOL is merely
COBOL rescaled. This is perhaps the best example of how the soccer
ball diagram reinforces traditional program thinking.

Another illustration of where the object difference might be observed: the

answer to the question of how many objects (classes actually) are required to
model a complete spectrum of typical business forms. Stated another way, how
many objects are required to create any possible business form? Your objects
will need the ability to collect a variety of information values (dollar amounts,
alphabetic strings, dates, and so forth) and make sure that each item collected
on the form is valid. When this question is posed as a class exercise, students,
traditional developers all, typically list 10 to 20 objects, including all kinds of
“managers,” “controllers,” and different “widgets” for each type of data to be
collected on the form (integerDataEntryField, stringDataEntryField, dateData-
EntryField, and so on).
An object thinking solution would tend to limit the number of objects to
four: a Form, acting as a collection; Strings, each label or text element being but
an instance of this single class; entryFields, a widget object that displays itself
and accepts input from a person; and selfEvaluatingRules, a special kind of
object that embodies a business rule or, in this case, a data value validation rule.
Both entryFields and Form objects would have a collection of selfEvaluating-
Rules that they could use to validate their contents.
 Chapter 1 Object Thinking 5

Note selfEvaluatingRules will be discussed in detail in Chapter 10.

For now, think of a self-evaluating rule as an expression (for example,
widget contents = integer object) in the form of an ordered sequence
of operators, variables, and constants. When the rule is asked to eval-
uate itself, each of its variables is asked to instantiate itself (become a
known value), after which the rule resolves itself (applies the operators
in the expression) to a final value—a Boolean true or false, for exam-
ple. A widget on a form is likely to have at least two rules, one to deter-
mine whether the entered value is of the correct class (for example, an
integer or a character) and one to determine whether the value is
consistent with the range of allowable values for that widget. If both of
these rules resolve to true, the widget can say its contents have been
validated.

Here’s a final example (for now) of places to observe the object differ-
ence: three different models of a customer. Figure 1-3a is an entity model: cus-
tomer as data. Figure 1-3b is a UML class, and Figure 1-3c is an object model.
Both Figure 1-3a and Figure 1-3b reflect typical thinking about customers as a
collection of facts that must be remembered by the system. Both examples
name the “object” and list its attributes. An attribute is the name of some char-
acteristic, the value of which must be remembered (stored by) the system. The
UML model (Figure 1-3b) differs from the entity model (Figure 1-3a) only
because it also lists operations—functions that manipulate the attributes—collo-
cated with the attributes themselves. The similarity between entity and UML
models tends to ensure that the latter are subject to the same rules of normal-
ization as entities. This means that attributes of an object tend to get dispersed
among several different objects, increasing the total number of classes created
to hold facts about customers.
Figure 1-3c models a customer only in terms of what it does—what ser-
vices it might provide. (The front-facing facet of the cube model contains a list
of these behaviors or responsibilities.) What is known (by the system or anyone
else) about the customer is not important to understanding the object itself.
What the customer object needs to know to satisfy its advertised responsibilities
is important and is evident (partly) in the left facet of the object cube depiction.
6 Object Thinking

Figure 1-3 Customer depictions.

F01MQ03

Entity modeling is most used in database applications where normaliza-

tion is required to build the underlying database properly. UML models need
not be normalized, but if you intend to fully utilize an automated tool, such as
Rational’s Rose, it becomes increasingly difficult to avoid normalization and the
explosion of classes that results from applying normalization rules. Object
models are seldom (if ever) concerned with normalization. The knowledge
stored in an object (leftmost facet of the model in Figure 1-3c) might be arbi-
trarily complex and need not be related in any formal way to whatever bit of
knowledge uniquely identifies the object.

Note There are advantages to modeling objects differently from enti-

ties. Here are two simple examples. First, it isn’t necessary to create two
different entities (classes) just because one object has an additional
attribute or needs to know an additional piece of information in compar-
ison with the other object, which reduces the total number of classes.
Second, it’s possible to aggregate a number of attributes into a single
collection of values (which we will later call a valueHolder). The object
has an instance variable named description, for example, which con-
tains a collection of discrete labeled values (dob:11/23/76, gender:F,
height:63, and so forth). The collection is itself an object with behav-
iors that allow it to add and delete label:value without the need to rede-
fine an entity and database schema and recompile the database.
 Chapter 1 Object Thinking 7

It’s not necessary to fully accept the Whorf-Sapir hypothesis (see sidebar)
to acknowledge that what and how we think varies with the language we
employ. Every mathematician knows the difficulty of translating elegant equa-
tions into English prose. Likewise, every programmer who has used more than
one programming language knows that statements easily and elegantly made in
one language (Smalltalk or Lisp, perhaps) are cumbersome and verbose when
translated into another (C++ or Java, perhaps). Developers are aware that dif-
ferent programming languages have very different idioms and styles despite
being, deep under the covers, equivalent.

The Whorf-Sapir Hypothesis

Benjamin Lee Whorf and Edward Sapir advanced a theory (overstated
here) to the effect that what you could and did think was determined by
the language you spoke and that some languages allowed you to think “bet-
ter” than others. Whorf was particularly interested in the Hopi language.
The Whorf-Sapir hypothesis is not fully accepted because most lin-
guists and cognitive scientists believe that the hypothesis is flawed by
overstatement and that the theory reflects some basic errors in the under-
standing of the Hopi language. However, most cognitive scientists do
believe there is an element of truth in the hypothesis. In recent years,
George Lakoff and others have advanced a similar theory with regard to
language and thought. Lakoff’s work centers on metaphor, a particular
aspect of a language, and shows how metaphor influences what and how
we think. Both Whorf and Lakoff are part of a school of philosophy and
cognitive science focused on the role of language as a shaper of thought.
The arguments for object thinking in this book reflect and are supported
by this kind of cognitive science research.

These simple examples, which I’ll return to and expand upon later in this
book, point to some of the differences resulting from object thinking, but other
more general comments can also be made about object differences. Consider
the continuum depicted in Figure 1-4. At the far left, the developer is thinking
about the problem and its intrinsic nature, and at the far right, the developer is
thinking of the most precise details of implementing a program. As we move
from left to right, we tend to think using different languages. At the far left, we
employ (or, since most of you are developers, you collaborate with someone
who employs) natural language and the vocabulary of domain experts. At the
far right, our vocabulary and syntax are essentially equivalent to an instruction
set embedded on a processor chip.
8 Object Thinking

Conception Execution

Decomposition Analysis Design Source Object

Code Code

Figure 1-4 Difference continuum.

F01MQ04

Object thinking occurs across the continuum illustrated in Figure 1-4, but
it’s far more critical when engaged in activities to the left of the spectrum than
to the right. In fact, if you apply object thinking correctly when you are concep-
tualizing and decomposing your problem space, you create a context that
frames and guides your thinking as you move to activities on the right half of
the spectrum. Object thinking and object decomposition yield a very different
set of objects (modules) than would result from using traditional decomposi-
tion thinking based on an understanding of data structures and algorithms.
Objects exist in the world. They are observed by and interact with other
objects—some of them human beings—in that world. Human objects come to
have expectations about the nature and the behaviors of other objects in their
world and are disconcerted if an object appears different or if it behaves in a
manner contrary to their expectations. The “naive” or “natural” expectations of
objects represent a kind of integrity—what Alan Kay called a user illusion—that
must be respected.
The set of objects produced by object thinking must eventually be
designed and implemented, ideally without violating the integrity of the objects
themselves and without contravening any object principles or ideals. This poses
a new set of problems for developers as they design classes and code methods.
As we move rightward on the continuum, thinking is constrained by the lan-
guages (remember Whorf and Sapir) that we must employ. Programming lan-
guages can lead us far from object thinking. Even if the language is relatively
benign (a “pure” OO programming language), it is still quite possible to write
nonobject code.
As I’ve said, although object thinking will yield observable differences
across the thinking continuum, the differences will be far more evident, and
their “correctness” will be essential, on the left. On the right of the continuum,
the differences will tend to be collective rather than individual. For example,
idiosyncrasies of syntax aside, it can be very difficult to differentiate a specific
object-oriented method from a similarly specific procedural function. But the
complete set of methods and the manner in which they are distributed among
the objects (modules) will be quite different than would be the case in imple-
mentations conditioned upon traditional computer thinking.
 Chapter 1 Object Thinking 9

Note I refer once again to COBOL programs to illustrate some differ-

ences. It was the case that a single COBOL program would encapsu-
late a bunch of functions, data structures, and some overall control
logic. Object orientation is not achieved, however, merely by breaking
up the monolith of a traditional COBOL program. Object COBOL
implementations will use a greater number of objects (each one a sep-
arate program), but absent object thinking, each little program will be a
mirror of the larger single program they collectively replaced: data
structures, functions, and control logic. Without object thinking, the
developer repeats old thinking habits on a smaller scale. One specific
example of the perpetuation of this kind of thinking is the treating of
data as something passive that must be protected, manipulated, and
managed. In COBOL, data is defined separately from all the functions
that operate on that data; it just sits there, and it is used (and fre-
quently corrupted) by functions elsewhere in the program. Uniting data
and procedure is far more complicated and difficult to do than is
implied by the soccer ball model (discussed earlier), which simply
shows the same old things stuffed in smaller containers.

In what other ways will object thinking manifest itself? What metrics will
be different? Dramatic and frequently claimed measures include the oft-cited
reductions in lines of code and the amount of lead time required to deliver an
application. Lines of code, for example, in a well-thought-out object application
will be at least an order of magnitude fewer (sometimes two orders of magni-
tude). This means that a 1 million–line program, written conventionally by
developers thinking like a computer, can be duplicated with object thinking in
100,000 lines of code or fewer. Time to delivery is reduced by at least 50 per-
cent and often by as much as 70 percent. Projects scheduled to take 2 years to
complete can be done in 8 to 12 months.
Dramatic results were routinely reported in the early days of object enthu-
siasm but have died off in recent years because, I would claim, of a return to
dominance of old thinking habits. This does not mean that object thinkers
reverted to form—merely that as the number of people using OO programming
languages and tools but not object thinking expanded dramatically, the collective
ratio of object thinking to traditional thinking became smaller. When entire teams
were object thinking and were almost certainly engaged in the rich iterative com-
munication practices advocated by XP, the results were dramatic. If less than 10
10 Object Thinking

percent of the team members are object thinking and the others are thinking
traditionally, the overall effort will reflect the thinking style of the majority.
Mark Lorenz3 and Jeff Kidd published a set of empirically derived metrics
for successful object projects. (Although their book is somewhat dated, metrics
derived from my own consulting experience—which continues to the
present—are consistent with those reported by Lorenz and Kidd.) Following is
a selection of metrics based on the results reported by Lorenz and Kidd but
modified by my own consulting experience and observations. The debt to
Lorenz and Kidd must be acknowledged, but they should not be held account-
able for my interpretations and extensions.

■ Application size: between 1 and 40 stories, with stories being roughly

equivalent to the story cards talked about in XP or one-fifth of the
use cases advocated in UML.
■ Classes per application: fewer than 100, which are a mix of key
classes (for example, business objects such as customer, account,
inventory item) and support classes (for example, collections, strings,
numbers, widgets). (See the sidebar "How Many Objects" in Chapter 4
for a discussion regarding the number of classes expected in a
domain class hierarchy.)
■ Number of classes required to model an entire domain (business
enterprise): approximately 1000.
■ Number of development iterations (conceive, develop, deploy,
evaluate—what Grady Booch called a “roundtrip”): three, on average.
Because XP iterates at the level of story and of release, this metric needs
to be reinterpreted. But some variation of three cycles per defined
effort seems consistent in both traditional and agile development.
■ Amount of code discarded at each iteration: 25 to 30 percent. This
metric is a proxy for the learning that took place and the refactoring
of code done at each iteration.
■ Responsibilities per class: average of seven. (We are using the term
class in this context as an exemplar object—so these responsibilities
are really object responsibilities, a distinction that will be made
clearer in Chapter 4, “Metaphor: Bridge to the Unfamiliar.” The same
is true when we talk about methods.)
■ Methods (the blocks of code that are invoked to fulfill a responsibility)
per class: average of 12.

3. Lorenz, Mark, and Jeff Kidd. Object-Oriented Software Metrics. Englewood Cliffs, NJ: Prentice Hall.
1994. ISBN 0-13-179292-X.
 Chapter 1 Object Thinking 11

■ Lines of code per method: average of fewer than seven for Smalltalk
(15 for languages such as C++ and Java).
■ Percentage of methods requiring inline (comments) documentation:
about 60. Practitioners of XP would think this a very high percent-
age. Refactoring, naming conventions, other idioms, and coding
standards should move this figure much closer to 0 percent.
■ Average method complexity using the following weights for each
occurrence of the noted item—API calls 5.0, assignments 0.5, arith-
metic operators or binary expressions 2.0, messages with parameters
3.0, nested expressions 0.5, parameters 0.3, primitive calls 7.0, tem-
porary variables 0.5, messages without parameters 1.0: average fewer
than 20 total points per method.
■ Number of case statements, nested conditionals, and nested branch-
ing statements in methods: zero (0). Lorenz and Kidd are quite care-
ful to cite a range for all of their metrics—a range that I have omitted
for dramatic purposes. The nature of the domain being modeled, the
application problem within the domain, and the relative skill of the
developers will affect the metrics cited. Treat the numbers as if they
were channel buoys—markers placed in the middle of the channel.
You do not have to hit the number, but sail too far to the side, and
you risk running aground.

Failing to achieve the metrics reported by Lorenz is a clear sign of the

absence of object thinking. In cases in which traditional thinking prevails, the
lines of code, the number of methods, and the complexity of methods will
increase dramatically. In cases in which thinking derived from database design
and data modeling prevails, the number of classes required to model any
domain and to build any application in that domain explodes.4 Another sign of
the ascendancy of traditional thinking over object thinking will be a plethora of
classes with the word controller or manager embedded in the class name.

4. One of the largest object projects of which I am aware through my consulting practice was tainted by
datacentrism. At the inception of the project, management mandated the translation of a recently com-
pleted enterprise data model into the domain class hierarchy—each entity in the data model became
a class in the object model. Despite using the right language, having some of the best object develop-
ers available in the country at that time, and having significant corporate support, the project took two
to three times as long to complete as necessary, cost far more than it should have, and encountered
numerous problems and complications that could have been avoided if the development team had
been allowed to rethink what they were doing using object thinking principles. Management deemed
the project “successful” because it was completed with no major overruns and saved the company a
lot of money. Compared to what might have been, however, it was a dismal failure.
12 Object Thinking

As the discussion of objects and object thinking proceeds, I’ll note many
other ways in which the object difference is observable. By the end of this
book, you should be able to clearly differentiate between object thinking and
traditional thinking solutions to problems. You should also be able to distin-
guish between “good” and “better” object solutions to a problem—recognizing
that design always yields a spectrum of potential solutions to any problem
and that evaluation of those solutions is partially subjective. Given that
observable and real differences exist between designs that result from object
thinking and from traditional thinking, what accounts for those differences? A
full answer will require the rest of this book, but I’ll introduce one essential
difference immediately.

Object Thinking = Think Like an Object

The essential thinking difference is easily stated: “Think like an object.” Of
course, this statement gets its real meaning by contrast with the typical
approach to software development: “Think like a computer.” Thinking like a
computer is the prevailing mental habit of traditional developers. And, as David
Parnas noted,5 it’s a bad habit:

The easiest way to describe the programming method used in

most projects today was given to me by a teacher who was
explaining how he teaches programming. “Think like a
computer,” he said. He instructed his students to begin by
thinking about what the computer had to do first and to write
that down. They would then think about what the computer
had to do next and continue in that way until they had
described the last thing the computer would do.… Most of
today’s textbooks demonstrate the same method, although it has
been improved by allowing us to describe the computer’s
“thoughts” in larger steps and later to refine those large steps
into a sequence of smaller steps.

This intuitively appealing method works well—on problems too

small to matter. … As we continue in our attempt to “think like
a computer,” the amount we have to remember grows and
grows. The simple rules defining how we got to certain points in

5. Parnas, David Lorge. “Software Aspects of Strategic Defense Systems.” American Scientist 73 (1985).
pp. 432–440.
 Chapter 1 Object Thinking 13

a program become more complex as we branch there from

other points. The simple rules defining what the data mean
become more complex as we find other uses for existing
variables and add new variables. Eventually, we make an
error.

In many of our computer systems there are several sources of

information and several outputs that must be controlled. This
leads to a computer that might be thought of as doing several
things at once. If the sequence of external events cannot be
predicted in advance, the sequence of actions taken by the
computer also is not predictable. … Any attempt to design these
programs by thinking things through in the order that the
computer will execute them leads to confusion and results in
systems that nobody can understand completely.

When there is more than one computer in a system, the

software not only appears to be doing more than one thing at a
time, it really is doing many things at once. There is no
sequential program that one can study. Any attempt to “think
like a computer system” is obviously hopeless.…

Programmers have tried to improve their working methods

using a variety of software design approaches. However, when
they get down to writing executable programs, they revert to the
conventional way of thinking.

It should be clear that writing and understanding very large

real-time programs by “thinking like a computer” will be
beyond our intellectual capabilities. How can it be that we have
so much software that is reliable enough for us to use it? The
answer is simple; programming is a trial and error craft. People
write programs without any expectation that they will be right
the first time. They spend at least as much time testing and
correcting errors as they spent writing the initial program. …
Software is released for use, not when it is known to be correct,
but when the rate of discovering errors slows down to one that
management considers acceptable.
14 Object Thinking

The kind of thinking described by Parnas in 1985 persists, even though

much new development is based on a data paradigm and even though most
people claim to be doing object-oriented programming. Developers’ thinking
patterns, their mental habits, have not changed. Programmers, and software
developers in general, still attempt to think like a computer.

Note Many readers will be familiar with most of the eminent scholars
and practitioners cited in this book but not necessarily all of them.
From time to time, it will be useful to introduce these authorities and
provide some background regarding their accomplishments, confirm-
ing that their insights do indeed contribute to the issues at hand.
“Behind the Quotes” will be a continuing sidebar used for this purpose.

Behind the Quotes

Dr. David Lorge Parnas
Dr. David Lorge Parnas, P. Eng., Ph.D., Dr. h.c., FRSC, FACM, Dr. h.c.,
Director of the Software Engineering Programme, Department of Comput-
ing and Software, McMaster University, is the Director of the Software
Engineering Programme in the Faculty of Engineering's Computing and
Software Department. He has also held nonacademic positions advising
Philips Computer Industry (Apeldoorn), the United States Naval Research
Laboratory in Washington, D.C., and the IBM Federal Systems Division.
The author of more than 200 papers and reports, Dr. Parnas is interested
in most aspects of computer system design. Dr. Parnas won an Association
for Computing Machinery (ACM) Best Paper Award in 1979 and two Most
Influential Paper awards from the International Conference on Software
Engineering. He is the 1998 winner of ACM SIGSOFT's Outstanding
Research Award. Dr. Parnas is a Fellow of the Royal Society of Canada and
a Fellow of the Association for Computing Machinery. He is licensed as a
professional engineer in the province of Ontario, Canada.
Dr. Parnas was selected to head the software portion of the Strategic
Defense Initiative (President Ronald Reagan’s program to develop a mis-
sile defense system, commonly known as “Star Wars”)—a project from
which he resigned when he and the other assembled experts concluded

(continued)
 Chapter 1 Object Thinking 15

Behind the Quotes (continued)

the software could not be written and tested. The quote used here is from
a paper he wrote analyzing the problem and the abilities of software
developers to address that problem. That paper was part of the public
explanation of his resignation.
Few individuals can match the stature earned by Dr. Parnas in the
realm of software development. Fewer still can match his moral conviction
and willingness to uphold principle over mercenary considerations.

Now, it might be argued that the computer itself has changed substantially
since 1985 such that the thing we are trying to think like is substantially differ-
ent. We can investigate this claim by looking at programming languages. A pro-
gramming language represents a virtual machine. Therefore, wouldn’t an
object-oriented language such as Java represent a very different kind of com-
puter? If we “think in Java,” don’t we “think like objects” instead of engaging in
the kind of “computerthink” discussed by Parnas? Unfortunately, no. As we will
see in Chapter 2, programming languages are shaped by their own set of pre-
suppositions. Although you can implement object thinking in any programming
language, knowing an “OO” language is not sufficient to engender object think-
ing. In Chapter 3, an example is presented that shows just how easy it is to rep-
licate traditional thinking in a “pure” object language such as Smalltalk.
Languages such as Java and object-modeling tools such as UML effectively
define an object as the encapsulation of data structures and algorithms. Accord-
ing to this kind of definition, an object (as discussed previously in this chapter)
is nothing more than a very tiny COBOL program. A COBOL program is the
encapsulation of data structures (data division) and algorithms that operate on
those data structures (procedure division). (In fact, creating object-oriented
COBOL was a relatively simple task involving making the program smaller in
scope and increasing the power of the interface division to make it easier for
one COBOL program to interact with others.) This example shows that devel-
opers tend to perpetuate old thinking habits, albeit sometimes to a lesser
degree.
It’s useful to generalize Parnas’s observation to accommodate datacentric
development and development in nonprocedural languages such as Prolog and
Lisp. The essence of thinking like a computer is thinking in terms of the means
of problem solution. You view every development problem through the lens of
the means you intend to use to solve that problem. In datacentric development,
the means of solution consists of relations and relationships; for Prolog and
LISP, a set of propositions or state declarations.
16 Object Thinking

The generalization being made here is a variant of the adage, “If your only
tool is a hammer, every problem looks like a nail.” Following Parnas, I’m sug-
gesting that software development has been distorted and that all the problems
noted by Parnas persist because “our only tool is a computer, so every problem
looks like a virtual machine.”
So, what is the object difference? How is thinking like an object different
from thinking like a computer? Throughout this book, I’ll be working to provide
a complete answer to this question, but I’ll preview a couple of differences
here:

■ Object thinking involves a very different means of solution—a coop-

erating community of virtual persons. A virtual person is an entity
capable of performing certain tasks. It has access to all the knowl-
edge and resources required to complete its assigned tasks. Both vir-
tual persons and virtual machines (computers) are really metaphors
for how we think about problem solution—but they are very differ-
ent metaphors.
■ Object thinking focuses our attention on the problem space rather
than the solution space. Object thinkers take the advice of Plato, Par-
nas, Fred Brooks, Christopher Alexander, and many others by letting
the problem define its own solution.

Problem = Solution
In Notes on the Synthesis of Form (Harvard University Press, 1964), Christopher
Alexander argues that “every design problem begins with an effort to achieve
fitness between two entities: the form in question and its context. The form
is the solution to the problem; the context defines the problem.” As an example
of “fitness,” Alexander points to the task of making a metal face smooth and
level. A standard block, smooth and level to a precision greater than required
by the target metal face, is inked and rubbed against the target. The presence or
absence of ink on the target indicates high and low spots that need to be
reground. The standard block is the context, the target face is the solution, and
it is clear how the standard defines the target.
I’ll develop the second point more fully in the next chapter, but again,
here’s a small preview. In their discussion of the principle of modular software
design, Witt, Baker, and Merritt 6 quote Plato and Christopher Alexander
(emphasis added):
 Chapter 1 Object Thinking 17

[First,] perceiving and bringing together under one Idea the

scattered particulars, so that one makes clear … the particular
thing which he wishes [to do] … [Second,] the separation of the
Idea into classes, by dividing it where the natural joints are,
and not trying to break any part, after the manner of a bad
carver … I love these processes of division and bringing together
… and if I think any other man is able to see things that can
naturally be collected into one and divided into many, him will
I follow as if he were a god.
——Plato, circa 400 B.C.

In practice problems are not homogenous. They are full of knots

and crevices which exhibit a well-defined structure. An
analytic process fails only if it does not take this structure into
account.
——Christopher Alexander, 1964

In these quotes, Plato and Alexander—and Parnas,7 Brooks,8 and Alex-

ander9 again, in other contexts—are talking about the notions of coupling and
cohesion, familiar terms to any software developer. However, they’re not talking
about the structure of a computer program; they’re talking about finding the
naturally occurring divisions and classifications (“natural joints”) in the problem
space. This is an essential difference because the naturally occurring modules in
a problem space are not isomorphic (exhibiting fitness between context and
solution) with the modules discovered and defined when you are thinking
about the design of a computer program.

6. Witt, Bernard I., F. Terry Baker, and Everett W. Merritt. Software Architecture and Design: Principles,
Models, and Methods. New York: Van Nostrand Reinhold. 1994. ISBN 0-442-01556-9.
7. Parnas, D.L. "On the Criteria to Be Used in Decomposing Systems into Modules," Communications of
the ACM, Vol. 15, No. 12, pp. 1053–1058, December 1972.
8. Brooks, Fred. “No silver bullet, essence and accidents of software engineering,” Computer Magazine,
April 1987.
9. Alexander, Christopher, Sara Ishikawa, and Murray Silverstein. A Pattern Language. Oxford University
Press. 1977.
18 Object Thinking

But they can be. Thinking like an object will lead to a greater degree of
isomorphism between objects found in the problem space (the enterprise
domain) and those employed in a solution space (the computer program) than
thinking like a computer. Isomorphism of the modules (objects) in problem and
solution space is a desirable, in fact essential, quality for software. I’ll demon-
strate the validity of these assertions in the coming pages.

Behind the Quotes

Christopher Alexander
Christopher Alexander is an architect. Not a software architect; rather, the
kind who designs buildings and cities. For an “outsider,” he has signifi-
cantly affected the practice of software development. I’ll cite the work and
ideas of Alexander frequently in this book as well as add biographical and
background information in later sidebars.
Alexander first came to the attention of the software community in
1968. He was frequently cited in papers at the NATO-sponsored first con-
ference on software engineering; that conference marked the widespread
adoption of software engineering as an appropriate metaphor for devel-
opment. The ideas on design and formal process that he articulated in
Notes on the Synthesis of Form were cited as appropriate for emulation by
budding software engineers. Many years later, his work on architectural
design patterns in A Pattern Language: Towns, Buildings, Construction
(Oxford University Press, 1977) inspired the software patterns movement
of the past decade.
Although most followers of Alexander’s ideas, and Alexander himself,
believe the software community missed the real essence of his ideas and
settled for simply copying a syntactic form, his influence is undeniable.

Object Thinking and Agile Development Practices

The second major objective of this chapter was to provide some initial argu-
ments for the special utility of object thinking to agile developers. To answer
this question, let’s look at the four XP values and a selection of the 12 XP prac-
tices introduced by Kent Beck.10
 Chapter 1 Object Thinking 19

Values
“We will be successful when we have a style that celebrates a consistent set of
values that serve both human and commercial needs: communication, simplic-
ity, feedback, and courage.” This statement introduces Kent Beck’s discussion
of values in eXtreme Programming eXplained and is a suitable introduction for
our own examination.

Communication
The communication value has two aspects: advocating greater communication
among all team members (developers, users, and managers) and promoting an
increase in the quality of communication. Object thinking directly contributes to
the latter aspect.
Imagine how much better your communication would be if all team mem-
bers spoke a common vocabulary and that vocabulary reflected a common
understanding of the problem domain and the essence of the problem being
addressed by the team. Imagine further that the concepts behind this vocabu-
lary reflected common understanding and did not require the customer side of
the team or the developer side of the team to memorize arcane terminology
employed by the other side.
Objects and object thinking promise the benefit of a common language.
Users, managers, domain experts, and developers will all use the same vocabu-
lary, grounded in a common understanding of that vocabulary, whether they
are discussing workflow or programming, requirements (stories) or refactoring. It
would be difficult to overstate the importance of a common language, especially
in an environment shaped by XP practices, almost all of which require deep
levels of communication among all participants in the development activity.

Note A domain expert is, by definition, someone with a deep under-

standing of the problem space and who has an ongoing involvement
with that problem space—a real live customer, in other words. A soft-
ware professional who has merely developed systems for “that kind of
problem area” before is not a domain expert. This is one reason why
XP demands an on-site customer.

10. Beck, Kent. eXtreme Programming eXplained: Embrace Change. New York: Addison-Wesley. 2000.
ISBN 201-61641-6.
20 Object Thinking

Simplicity
Every aspect of object thinking shares a common goal with this XP value, which
equates to finding the simplest thing that could possibly work. Some examples:

■ Object identification and responsibility assignment are based on the

structure of the problem domain rather than on the structure of some
potential computer program–based problem solution. However
complex or complicated the natural world appears, human beings
do very well in coping with that world. This suggests that a kind of
natural simplicity is implicit in that world and is waiting to be discov-
ered and employed.11
■ Object thinking will lead to the smallest number of distinct things
(represented as classes) possible—a kind of numerical simplicity.
■ Objects designed and constructed following the application of object
thinking will do the least amount of work possible and do it in the
most direct and simplest way possible.
■ Focusing on coordination of autonomous objects instead of the con-
trol and management of unruly modules and passive data structures
will also contribute to simplicity.

Feedback
Object thinking’s contribution regarding feedback is somewhat indirect. Object
thinking evolved, in part, from the rapid prototyping school of software devel-
opment and shares the values of immediate feedback that arose there. Object
tools, such as the Smalltalk programming language, were optimized to provide
immediate feedback to developers. Both the use of an interpreted language and
the provision of multiwindowed development environments with browsers and
workspaces are manifestations of this desire for rapid feedback. Consider the
importance of the kind of immediate feedback described by Michael Hiltzik
when talking about Apple’s famous visit to Xerox PARC:

Given this rare psychic encouragement, the Learning Research

Group warmed to their subject. They even indulged in some of
their favorite legerdemain. At one point Jobs, watching some
text scroll up the screen line by line in its normal fashion,
remarked, “It would be nice if it moved smoothly, pixel by pixel,
like paper.”

11. Christopher Alexander’s current works on the nature of order (http://www.patternlanguage.com) are
a quest to find this natural—but exceedingly powerful—simplicity.
 Chapter 1 Object Thinking 21

With Ingalls at the keyboard, that was like asking a New

Orleans jazz band to play “Limehouse Blues.” He clicked the
mouse on a window displaying several lines of Smalltalk code,
made a minor edit, and returned to the text. Presto! The
scrolling was now continuous.
The Apple engineer’s eyes bulged in astonishment.

Courage
Object thinking embodies precisely the kind of courage advocated by Kent
Beck:

Every once in a while someone on the team will have a crazy

idea that just might slash the complexity of the whole system. If
they have courage, they’ll try it out. It will work (sometimes).

Object thinking is a “crazy idea” capable of increasing simplicity in software

design—crazy in the sense that it does not conform to traditional thinking about
software development, crazy in the sense that it revolts against the computer
thinking employed by most software developers, and crazy in many of the
same ways that XP is crazy. It takes courage to take a stand in opposition to
conventional wisdom as to how software should be developed. It takes courage
to suggest that formal practices, scientific practices, and engineering practices
are not necessarily the best way to develop software and that we might learn
something from philosophy, anthropology, and art.
Courage is not just “crazy ideas” and the ability to take advantage of them.
Courage is the ability to confront and overcome fear:

■ Fear that changes will break your software. Properly constructed

objects, enforced encapsulation, and diligent refactoring based on
solid object thinking reduce the risks associated with change to a
negligible level.
■ Fear of mistakes. Mistakes are simply one iteration in the learning
cycle. Object thinking is grounded in the idea that exploration and
experimentation are intrinsically valuable and that learning what
doesn’t work is just as important as learning what does work. There
are no mistakes: every result tells you something of value about what
you are trying to accomplish.
22 Object Thinking

■ Fear that change will be costly. Change is costly only if the effects of
a small incremental change in one part of a system are propagated
throughout the system and somehow precipitate widespread failures
in other parts of the system. Objects and object thinking assure that
change is a local, not a systemic, phenomenon.

Selected Practices
Although it would be possible to link object thinking with all of the practices
advocated by XP, a few merit explicit mention.

Metaphor
Metaphor is one of the most important but least understood practices of XP.
Some12 would claim that all thinking is metaphor-based. Kent Beck spent his
OOPSLA 2002 keynote talking about the importance of metaphor. Object think-
ing introduces a very different set of metaphors to use in talking and thinking
about software development. Object thinking metaphors are compatible with
XP and are an important complement to XP metaphoric thinking. The common
language provided by objects and object thinking provides a system of meta-
phors: every story, every test, every code module employs labels and terms that
metaphorically connect things in the domain with their simulated coded coun-
terparts.

Simple Design
Object thinking leads to simple design (usually the simplest design possible);
the least number of classes necessary; the fewest number of methods per class
possible; the simplest coding of methods; the avoidance of control, centraliza-
tion, and management classes; and simple scripts to simulate simple stories.

Refactoring
Refactoring is a way for “lazy” objects to give all the hard work to other objects.
Of course, when refactored and distributed to the proper objects, the work
turns out not to be hard at all. Whenever a task looks too hard for a good
object to perform by itself, it looks for others to share in the workload. Succes-
sive passing of hard work to others results in no object doing anything really
difficult—and usually in a collective solution that is intrinsically simpler than
would have been possible if the work had remained where originally assigned.
For example, suppose an airplane object has a responsibility to report its
location. This is a hard task because the location is constantly changing; a

12. George Lakoff, for one. His work on cognition represents a fairly dramatic shift in this area of study,
and his findings are very important for software developers.
 Chapter 1 Object Thinking 23

location is a composite structure (latitude, longitude, altitude, direction, speed,

and vector); the values of each part of that structure come from a different
source; and someone has to remember who asked for the location and make
sure it gets back to them in a timely fashion. If the task is broken up so that

■ The airplane actually returns a location object to whoever asked for

it after appending its ID to the location so that there is no confusion
about who is where. (We cannot assume that our airplane is the only
one reporting its location at any one time.)
■ An instrument cluster keeps track of the instruments that must be
asked for their current values and knows how to ask each one in
turn for its value (a collection iterating across its contents).
■ An instrument merely reports its current value.
■ A location object collects and returns a set of label:value pairs (alti-
tude:15,000 ft.).

None of the objects do anything particularly difficult, and yet collectively

they solve a complicated problem that would be very hard for any one of them
to accomplish individually.

On-Site Customer
The value of the on-site customer increases when she can fully participate in
the development work by virtue of the common vocabulary described earlier in
the “Communication” section. Object thinking not only creates a common
vocabulary but also allows the on-site customer to actually make design sugges-
tions and validate emerging designs as powerful representations of business
context and goals because the behavior of objects is simple, transparent, and as
familiar to the user as it is to the developer.

Coding Standards
Let me quote Kent Beck again: “Programmers write all code in accordance with
rules emphasizing communication through the code.” Object thinking pro-
motes this kind of coding. In fact, much of what will be presented on coding
later in this book is strongly influenced by Beck’s Best Smalltalk Practice Pat-
terns (Prentice Hall PTR, 1996), a style book for good coding. The coding metrics
noted earlier—lines of code per method and number of methods requiring
inline comments—are consistent with this XP practice.
As in the rest of this introductory chapter, I’m making assertions and
claims about object thinking—here specifically about object thinking’s compat-
ibility with XP—that will be proved (or not) in the chapters that follow.
24 Object Thinking

Thinking Is Key
There was a time when software development was considered an art. Art
implies individual talent based on innate skills: “Artists are born, not made.” As
computers became increasingly prevalent and essential to modern industrial
life, the dependence on artists to create the software that made them useful was
deemed undesirable. Some were even offended that a computer—the embodi-
ment of rationality, mathematics, and hard science—could be corrupted by
“mere” art. Increasing demand for programming and development services sug-
gested that the pool of true artists was too small; some other way would have
to be found to create the mass of individuals needed to fill all the jobs available.
In the early sixties, a movement to eliminate art from development came
to dominate the way we thought about the manufacture (the terms educate and
train were used, but only as euphemisms for manufacture) of software devel-
opers. Tools would automate most of the development tasks, defined method
would replace idiosyncratic problem solving, and documented process would
allow every step to be monitored, measured, and controlled. Even the most
mediocre human raw material could generate superior results if all the tools,
methods, and processes were deployed properly.
The preface to this book begins with quotes from Robert L. Glass that sug-
gest that the attempt to eliminate humans—or at least humanity—from software
development failed. Humans and human abilities are still the key to software
development success. This, in turn, suggests that we need to reconsider how
we go about enhancing human abilities instead of attempting to replace them
with machine capabilities.
One small step in this effort is to reconsider the “artists are born, not
made” dictum. Thousands of art schools (ranging from small commercial acad-
emies to graduate academic programs) claim to make artists. While it may be
true that no school “creates” artists on the order of Michelangelo or Georgia
O’Keefe quality, they do enhance the innate talents of individuals and make
them the best artist they can be. The education of artists is not focused on tech-
nique, process, or tools. The majority of an art education combines ideas, his-
tory, appreciation, experience, and constructive criticism.
An artist who is aware of the philosophy and history behind the practice,
an artist who is attuned to the community in which art is created and appreci-
ated, an artist who is capable of deep communication, is a better artist.
This book believes the same thing to be true of software developers—soft-
ware artists, software professionals. Mastery comes first from a thorough under-
standing of ideas. Understanding of ideas requires an understanding of
context, including historical context. Mastery is shared—it is a property less of
the individual than of the group—and is shared by those participating in a com-
mon culture.
 Chapter 1 Object Thinking 25

Software Development Is a Cultural Activity

To succeed, a software developer must be able to comprehend enterprise
domains and perceived problems within those domains, conceptualize alterna-
tive configurations of artifacts and processes, and articulate ideas in a number
of languages—at a minimum, a natural language, a modeling language, and a
programming language. Because all of these activities involve thinking, improv-
ing a developer’s cognitive abilities will result in improvement in the software
he or she produces. This would seem obvious, but little has been done in the
world of computing and software development to directly address this issue.
Instead, the majority of software improvement efforts have focused on lan-
guages, tools, methods, and processes rather than on how to think about soft-
ware.
In fact, efforts to reform software development by introducing a new way
of thinking about software or about development are usually met with signifi-
cant resistance. More often than not, new ways of thinking are dismissed out of
hand or are co-opted and corrupted until they are acceptable to mainstream
developers. At that point, they are but pale imitations of the original ideas.
An observer of these conflicts over the past 50 years cannot help but be
struck by the similarity between the language wars, formalism wars, and
method wars occurring within the software development community and ethnic
and religious conflicts in the world at large. In fact, our conflicts over how to
think about software development reflect real, but usually tacit, cultural differ-
ences—the most obvious example being tacit assumptions about the value of
rationalism.
We live in a culture in which the value of, and the esteem given to, science
and rationality are so deeply embedded that we seldom consciously consider
why we have this pervasive bias. We do not question how and why this bias
became entangled in other aspects of our culture—gender, for example, where
a traditional association of rationality with the male gender and irrationality
with the female was used as a justification for perpetuating unequal treatment
based on gender. This same bias in favor of rationality is reflected in debates
over method—the implicit assumption is that a method is weak unless built on
a formal (rational, logical, mathematical) foundation. And it shows up in
debates about programming languages, which not only must be formal but
must resolve every hint of potential ambiguity, even to the point of adding fea-
tures capable of resolving conflicts that are unlikely to be encountered by any
programmer anytime in the next thousand years.
26 Object Thinking

Note I’ll say much more about cultural, and usually tacit, assump-
tions in the next two chapters.

Five significant software development innovations and trends have gained

prominence in the past 50 years: datacentrism, software engineering, object ori-
entation, patterns, and extreme programming. (Although computer-aided soft-
ware engineering [CASE], total quality management [TQM], and capability
maturity model [CMM] have been prominent innovations in software develop-
ment, they were not, and are not, primarily new ideas about how to think about
software.) Datacentrism—that is, focusing on data entities, relations, and data-
bases instead of algorithms, flow charts, and data-flow diagrams—has been the
most successful of the five thinking innovations, although most developers still
think procedurally when writing code. Both software engineering and object
orientation have achieved a strange status—everyone claims to be doing them
without really doing so. The patterns movement exists mainly as a buzzword,
and the jury is still out on extreme programming.
Antipathy toward new ideas, especially ideas that represent radical chal-
lenges to the perceived wisdom of formal methods and formal processes, con-
tributes to the demise of radical new ideas. But so too does the failure to
articulate radical ideas as ideas. Too often, proponents present their ideas in a
guise supposedly more acceptable or easier to communicate—for example,
objects in terms of a programming language (Smalltalk) or interface design as a
set of standards, or extreme programming in terms of being a method. This
kind of presentation almost always weakens the force of the new idea, making
it even more difficult to overcome opposition and avoid co-option and dilution.
Extreme programming—and to a great extent all of the agile methods—
represents a new and creative idea about software development (that is, an
idea based on old and proven practices, reformulated, and taken far more seri-
ously). The essence of this new idea might be distilled to a single sentence:
“Software development is a social activity.” Of course, it will take an entire
series of books to explicate this simple idea and all of its implications. But the
ultimate success of XP is dependent on people understanding and accepting
the idea, and the body of thought presupposed by and supporting that idea,
rather than on simply adopting the “twelve practices.”
 Chapter 1 Object Thinking 27

Behind the Quotes

Ward Cunningham and Kent Beck
Ward Cunningham is an almost mythical figure in the world of objects and
extreme programming. Excepting a few papers, usually coauthored by
Kent Beck, he has published little. His influence, however, has been mon-
umental. He is considered to be the inspiration for most of the extreme
programming practices, is a legendary coder and designer, and is a great
mentor. Mathematicians use an “Erdos Number” as an indicator of their
association with Paul Erdos, one of the most prolific and brilliant mathe-
maticians of the past century. Erdos himself had the number 0, those who
coauthored a paper with him had the number 1, coauthoring with a coau-
thor yielded 2, and so on. Extreme programmers are given a “Ward Num-
ber” based on pair programming with him (1), pair programming with
someone who paired with him (2), and so forth.
Kent Beck is to Cunningham as Plato was to Socrates—the extender
of ideas, contributor in his own right, and, in an interesting parallel, the
one who published. The preeminent figure in the world of extreme pro-
gramming today, Beck was equally active in object programming.
Although he did not publish a book on behavioral object modeling, he did
publish works on programming style and idiom (for Smalltalk).
Kent Beck and Ward Cunningham were a team when objects first
became a hot technology, collaborating both as developers and as creative
thinkers about object technology. Both will become quite familiar to the
reader, as their ideas are central to many of the themes in this book.

Implicit in the idea of XP are other supporting ideas. Notable among these
is the idea of objects. This is not surprising given that XP gurus Ward Cunning-
ham and Kent Beck played a central role in the object revolution. They are the
inventors of the CRC (Class-Responsibility-Collaborator) card approach to
object analysis. The kind of object thinking introduced by Cunningham and
Beck was, and is, far more consistent with the original object idea than any of
the other popular approaches. The majority of developers claiming to be fol-
lowing object-oriented principles use the approach popularized by Rational
Software under the umbrella of UML with nuances introduced by the Java
programming language. Although Grady Booch, one of the principal contribu-
tors to UML thinking, clearly understood and advocated the object idea in his
early work, little of that understanding was codified into UML.
28 Object Thinking

The CRC Card Approach

The project that brought object programming into the mainstream was the
development of Smalltalk at Xerox PARC in the 1970s. The first ideas
about discovery and specification of objects arose in that context—Object
Behavior Analysis (OBA) was the method developed by Adele Goldberg
and Ken Rubin. This method was never published—except for one brief
paper—but despite the flaw of overzealous formalism, it created part of
the foundation for behavioral object thinking.
Ward Cunningham developed the foundation for what was to
become a simple and powerful mechanism for capturing and understand-
ing object behavior—a simple 3 × 5 index card. The index card provided
a means for exploring and documenting a development team’s under-
standing of objects. The format of the card was simple: a class name
appeared at the upper left corner, and two column headers, Responsibility
and Collaborator, appeared on the next line of the card. Class, Responsi-
bility, Collaborator (CRC) headings yielded the popular name for the cards
and the approach to analysis and design based on those cards.
Figure 1-5 shows part of a completed CRC card for an airplane object.

Figure 1-5 A partial completed CRC card for an airplane.

F01mq05

Chapter 6, “Method, Process, and Models,” will introduce more detail

about CRC cards and how they became the foundation for the “Object
Cube” at the center of applied object thinking and modeling introduced in
this book.

Object thinking is central to extreme thinking. This is true even though

there is little explicit mention of objects in the XP books. Object thinking is so
basic to XP that it is simply assumed. As with cultural practices, it seldom occurs
to members of a given culture that there is any need to explicitly talk about
 Chapter 1 Object Thinking 29

those practices—you simply do them the way everyone does. Many of the XP
values and practices—for example, simplicity, stories, and refactoring—are
dependent in important ways on object thinking.
XP values and practices aren’t based on just any flavor of object thinking.
Object thinking in XP is based on the behavior-centric ideas about objects
exemplified in the CRC card approach. This approach to object thinking was
always a minority view and almost disappeared as UML attained dominance
and C++ challenged Smalltalk. (Visual Basic and Java superseded C++ in popu-
larity and offered an improvement for object thinking, but still not an optimal
environment.) Minority status was almost assured because of the close associa-
tion of the Smalltalk programming language with behavioralism, an “informal,
human-centric method” in a culture dominated by a bias in favor of formalism
and engineering.
Oft-cited reasons for the demise of Smalltalk and the behavioral ideas
behind informal methods (such as CRC cards) include performance issues, scal-
ing issues, and hyperbolic promises by expensive consulting companies lead-
ing to dramatic failures. These reasons appear to be little more than
rationalizations since every innovation in software has been subject to the same
criticisms. If an innovation can claim a formal foundation (relational databases),
it’s given time to overcome its limitations (or those limitations are ignored, and
the innovation is used in spite of them).

Note The same forces that are bringing XP to the forefront of soft-
ware development have generated a resurgent interest in the behav-
ioral approach to object thinking. In addition to this work, Rebecca
Wirfs-Brock and Alan McKean have published Object Design (Addi-
son-Wesley, 2003), and Richard Pawson and Robert Mathews have
published Naked Objects (John Wiley & Sons, 2002), both of which
strongly advocate and update behavior-based object thinking.

The purpose of this book is to help you become a better software devel-
oper and a better agile developer by improving the way you think. Thought,
however, is not simply a matter of rote memorization and repetition of a few
facts; it requires an understanding of ideas, of metaphors, and of history, topics
that many readers will find unusual in a computer book. Changing your mind
requires that you change your worldview, your culture, to include not only new
ideas but new values and new perspectives. This, in turn, requires a temporary
suspension of your current mindset. You must be prepared to take objects and
30 Object Thinking

object thinking seriously. An important aspect of taking the object idea seri-
ously is a willingness to change your mind, to give up (at least temporarily) old
and sometimes cherished ideas that are in conflict with new ones. Many of the
original proponents of object ideas overdramatized the need for a mental
change. (“Step one in the transformation of a successful procedural developer
into a successful object developer is a lobotomy.”) Predictably, this alienated
potential converts to the new ideas more often than it convinced them. I’ll try
to avoid this error in this book, while at the same time emphasizing those occa-
sions when you really must think differently in order to think in an object-ori-
ented fashion.
Of course, the object idea is not a single idea. It’s a complex of interrelated
ideas that provide a kind of philosophy or, perhaps more accurately, a mental
perspective. This perspective shapes the software developer’s thoughts con-
stantly and pervasively whenever a decision needs to be made, a problem
needs evaluation, or a design requires formulation. It matters little whether the
actual development act involves coding, testing, analysis, or design—the object
perspective influences that mental activity.
Object thinking will be most effective only after the developer internalizes
the object perspective. The same thing can be said about extreme program-
ming. In both cases, the novice is confronted with an array of terms, of prac-
tices, of examples that constitute a “gate” (to use Christopher Alexander’s and
many a Zen mystic’s terminology) that must be “passed through” to attain full
mastery. As I introduce vocabulary, models, and examples, remember that
these are but expressions of the idea (the gate), not the idea itself.
My excursion into history and philosophy in this book is intended to
expose the cultural foundations of our thinking about software development. It
will accomplish, hopefully, two things: expose the origins of divergent schools
of thought (XP and objects on one side; traditionalist computer scientists and
software engineers on the other) and provide the cultural context necessary to
fully assimilate the material in later chapters.

Onward
In this chapter, I tried to make a case for reading the rest of this book. Some
promises have been made, including that object thinking will make you a much
better software developer and that object thinking will make you a much better
XP practitioner.
 Chapter 1 Object Thinking 31

The journey toward mastery of object thinking is not direct. I will ask you
to take a short detour into history and philosophy in order to develop the con-
text from which object thinking emerges. This context should make it easier to
understand the advocacy of behaviorism as well as why terms and concepts are
developed as they are in later chapters.
A secondary, and in many ways more important, reason for looking at phi-
losophy and history is to establish in the reader’s mind the themes and princi-
ples that can help you expand and shape the details of object thinking to
accommodate new situations. It’s my hope that you learn object thinking not by
rote but by actively and creatively applying ideas, concepts, and philosophical
principles.
Object thinking predates agile thinking, and both share deep common
roots. Object thinking and agile development, especially extreme program-
ming, are natural partners, sharing values and following common practices.
Understanding these commonalities at the level of ideas and values is essential
to the mastery of both. Exploration of these similarities will allow you to
develop synergies of thinking and to become more agile.
Welcome to the journey.

Forward Thinking
Introduction
The first five chapters of this book introduce a lot of ideas and back-
ground information that are fundamental to object thinking. As critical as
these ideas are, many readers will find themselves anxious to get to the
material that illustrates application of the ideas. This continuing sidebar—
an example application problem—will partially satisfy the urge to get to
the “good stuff” while providing illustration of the ideas presented in early
chapters.
Be forewarned: the example may involve concepts and models that
have not been introduced in the main text. You may have to take a few
things on faith until they are discussed properly.

(continued)
32 Object Thinking

Forward Thinking (continued)

Problem Statement:
The Universal Vending Machine Project
It’s your first big XP project—you are to head a team building the software
for a “better vending machine.” Vending machine sales are flat, but your
CEO notes that in Japan almost anything can be purchased from vending
machines, and in Scandinavia people can buy products from machines
using their cell phones. The hardware side of the company is busy design-
ing and building customized vending machines of all types, and your team
is to develop a common software base to run every type of vending
machine they might come up with.
The Universal Vending Machine (UVM) will be capable of dispensing
liquids as well as packages (cans or products in wrappers of various
kinds). Payment can be made using any or all of three kinds of currency
and coins (U.S. dollars, euros, and yen), debit cards (including a line of
prepaid debit cards sold by your company), and credit cards. Customers
can purchase goods from the machine via a Web-based transaction—also
using debit and credit cards. Each vending machine will initiate product
reordering via the Web. Restocking will be adjusted based on demand—
which products sell the most in the least amount of time.
To save money, your team will create a single program, one capable
of supporting the entire line of planned UVMs.
Given that this is an XP project, your customer will provide you with
stories, and you will write tests and code programs. Getting from stories to
code requires a fair amount of thinking, some open discussion, definition
of tests, occasional writing of notes, and drawing of diagram fragments.
Fortunately, you and your team are conversant in object thinking, so this
will not present an insurmountable challenge.
Philosophical Context
More often than not, the first question programmers ask when embarking upon
a new development project is, “What language will be used for implementation?”
There is usually no need to ask about method because some variation of a formal
structured approach is assumed. This is a very unfortunate situation.
The rising popularity of extreme programming (XP) and agile develop-
ment makes the method question an open one—again. I say again because in
terms of observable actions, XP and agile approaches are just the latest incarna-
tion of iterative development. Some form of iterative development has been
practiced in software since the 1960s and has usually been held to be superior
to the structured “waterfall” approach that is officially practiced. And yet, when
XP was proposed about four years ago, management and most practitioners
reacted as if it were a radical departure from accepted and acceptable practice.
Later in this chapter, we’ll look at the philosophical context behind this reac-
tion. But first it’s useful to explore how philosophy shapes the development
and utility of programming languages.
Whatever the answer to the programming language question, it is almost
always given for the wrong reasons. Even worse, it is given for reasons that are
never articulated and therefore never subject to reasoned judgment. Common
reasons (not necessarily expressed out loud) for adopting a programming lan-
guage include, in no particular order, the following:

■ Loyalty “We are a Microsoft shop; we use Visual Basic (or, today,
Visual C#).”
■ Bandwagon “Everyone is doing Java.”
■ Economics “Java programmers are a dime a dozen and completely
interchangeable—if we lose one, we can find a replacement easily.”

33
34 Object Thinking

■ Culture “You can’t do telecom/real-time/embedded applications

in anything except C++.”
■ Resumé “All the job ads ask for C# experience, so I had better get
some exposure.”
■ Inertia “I wrote my first program in COBOL and am most familiar
with COBOL, so COBOL is the right language for this project. Using
COBOL makes it easier for me to manage the project.”

Given that all programming languages are ultimately equivalent—“anything

you can do, I can do also”—are there any legitimate reasons for selecting one
language over another? I would suggest three:

■ Pragmatics Mature libraries, compatibility with external systems

(if you want to be a supplier to Wal-Mart, you make your systems
compatible with theirs), availability of labor force, or other economic
reasons. However, my professional experience suggests that 70 to
80 percent of the pragmatic cases made for a particular language
have been rationalizations, not rationales. The pragmatic case has
usually been a surrogate for loyalty, bandwagon, and similar reasons.
■ Performance If, and only if, performance mandates cannot be
satisfied with effective design, it might be appropriate to select a lan-
guage that provides more direct access to and control of hardware.
Here again, performance is more often than not an excuse and not a
reason. It’s possible to build hard real-time embedded systems using
“slow and cumbersome” languages such as Smalltalk without paying
any performance penalty compared with C or C++ implementations.
It takes excellent design and the occasional programming trick (pre-
compiling methods, for example), but performance is rarely a true
reason for selecting a particular implementation language.
■ Philosophy Programming languages are created for a reason. Lan-
guage designers have specific intentions for their language, and
those intentions are grounded in and shaped by particular philo-
sophical values and ideas. Developers and development teams
should select an implementation language that has been shaped
by philosophical ideas and ideals similar to their own and to their
problem space.

The philosophy implicit in the implementation language should be the

reason for selection. There are some problems with putting this dictum into
practice. Most developers, especially programmers, might be quite surprised
 Chapter 2 Philosophical Context 35

that philosophy plays any role, let alone a critical role, in language design
because few have had the occasion or opportunity to explore the ideas behind
languages. Most object developers are not conversant with the philosophy of
objects—the subject of this book—and are therefore in a poor position to select
an implementation language reflective of object philosophy.
Developers who specialize in software for large-scale switching networks
(such as the phone system) should choose a language such as C because the
ideas and ideals that shaped that language are the same ones that shape the
thinking of developers working in that problem domain. (Even C++ will be
seen as an intrusion by those developers.) Scientific or engineering number
crunching? No reason not to use FORTRAN. Business applications? COBOL or
Visual Basic is perfectly fine.
Philosophical compatibility will allow you to be more expressive than if
you are trying to translate your thoughts into a “foreign” language because your
thoughts will naturally flow into the syntax of the implementation language.

Note Every programmer has a first language—the one she learned

first and used most. Most programmers have occasion to learn addi-
tional languages during their careers. Languages can be classified
into families like procedural (C, COBOL, PASCAL) and declarative
(Lisp, Prolog). Learning a new language within a family is relatively
easy (Java in 40 hours, if you learned Pascal first; however, learning a
language across families is quite difficult (C programmers learning
Lisp). Object languages like Smalltalk proved most difficult for expert
C programmers and expert database developers precisely because it
required them to cross language family lines. Unless the programmer
also learns a great deal of idiom and convention (not included in the
formal syntax of the language or even in its libraries), she will continue
to write (for example) COBOL using Java syntax.

Developers might, initially, be surprised at the significant and often hid-

den role played by philosophy. Many, if not most, have experienced moments
of frustration when a language provided no direct and obvious way to express
the ideas they had in their heads. Such moments are frustrating precisely
because the true source of friction is hidden—it’s based on unstated philo-
sophic incompatibilities. On the other hand, many programmers have experi-
enced the exhilaration that comes when design thinking and implementation
language are compatible and productivity soars. (Users of script languages for
Web development are likely to enjoy this type of experience.)
36 Object Thinking

On a macroscopic level, it’s impossible to truly understand the acrimony

and the bitterness of the debates about object programming languages without
looking beyond the languages themselves to the philosophy implicit in those
languages. In the case of object language debates, the philosophic problem is
twofold: first, very few people have explored the philosophy behind the lan-
guages claiming to be object-oriented (even though that history is well docu-
mented); second, few have understood object philosophy, making it difficult to
match that philosophy with the one behind a given programming language. A
short foray into the history of three languages of particular interest to object
thinkers—SIMULA, C++, and Smalltalk—will illustrate the problem.

Philosophy Made Manifest—Dueling Languages

Computers—actually the engineering behind the construction of various hard-
ware devices—and mathematics are the binary stars around which the world of
computer science has revolved for more than fifty years. Programming once was
a matter of rewiring. It was a decade before machine-level problems were solved
to the extent that we could turn our attention to larger issues of program design.
Another decade elapsed before we worked on analysis of applications and sys-
tems employing computers and programs. Significant attention is now being paid
to the “soft” issues of usability, human/machine interaction, and even culture.
Despite these advances, whenever a new idea is introduced in computer
science, it receives little attention until it has been made concrete and explicit—
frequently as a new programming language1. Debates about ideas are then
transformed into debates of the relative merits of the artifact languages. Argu-
ments about design, analysis, method, and process inevitably follow, but those
discussions are more often grounded in the programming language artifacts
than in the original ideas. In this process, the original ideas become secondary
if they are not lost entirely.
This has certainly been true in the case of the “object idea” and the program-
ming languages that lay claim to that idea. The most vituperative debates in the
object community have tended to center on questions of programming language.
The overt focus of those debates has been on programming language features and
technical benchmarks. But it is the emotions induced by covert (not consciously
hidden, merely assumed and unspoken) philosophical positions that have
accounted for the intensity and the hostility evident in those debates.

1. All of the quotations regarding SIMULA, Smalltalk, and C++ languages are taken from the respective
chapters in ACM History of Programming Languages, ACM Press, 1978, Richard Wexelblat (ed.), or
History of Programming Languages II, Addison-Wesley, 1996, Thomas J. Bergin Jr. (ed.).
 Chapter 2 Philosophical Context 37

Until the appearance of Java and C#, Smalltalk and C++ were the prime
contenders for the hearts and minds of object developers. The intensity of argu-
ment between advocates of each language is legendary. Both sides tended to
see Java as an interloper and tended to criticize those aspects of Java that
reflected their more traditional nemesis. Disagreement between Smalltalkers
and C++ers gains added interest from the fact that both claim to be the direct
heirs of another, older, language, SIMULA.

Behind the Quotes

Alan Kay, Kristen Nygaard, and Bjarne Stroustrup
Each of these individuals is known for designing and promoting a partic-
ular programming language (which we will be discussing shortly). Each
individual is also known for many other contributions to the world of soft-
ware development and computer science.
Alan Kay designed, and Dan Ingalls implemented, the Smalltalk
programming language while working at Xerox PARC. After leaving Xerox
PARC, they both worked at Apple Computer, where they reinvented Small-
talk and called it Squeak. Squeak underwent further development when
both men moved to The Walt Disney Corporation. Today they are inde-
pendent developers focused on educational applications of the Squeak
language and its multimedia capabilities.
Dr. Kay, who received his Ph.D. in computer science from the Uni-
versity of Utah, developed the idea of a Dynabook—a network-enabled
laptop computer with the Smalltalk environment—also while at Xerox
PARC. Arguably, the Pad Computer version of the Microsoft Windows
operating system would be—if coupled with Smalltalk/Squeak—a pretty
close approximation of his ideas.
Kay’s biggest contribution was changing the way industry and users
think of computers. Before Kay, computers were impersonal tyrants that
required you to speak their language and limited their interaction to text-
based communications. Kay personalized the machine: he expanded the
means of interaction to include graphics, point-and-click, and multimedia
using a language designed for humans rather than machines. The PC on
your desktop reflects more of Kay’s ideas than most people realize.
Kristen Nygaard developed, with Ole Johan-Dahl, the SIMULA pro-
gramming language, introducing the concepts upon which all later object-
oriented programming languages are built: objects, classes, inheritance,

(continued)
38 Object Thinking

Behind the Quotes (continued)

virtual quantities, and multithreaded program execution. He was also a

social activist greatly concerned with the use of information technology
and spent a good part of his career concerned with the social impact of
computer technology. His work became the foundation of what today is
called the Scandinavian school of system development, closely linked to
the field of participatory design.
Nygaard has received the Norbert Weiner prize from Computer Pro-
fessionals for Social Responsibility (CPSR) for responsibility in social and
professional work; the Rosing Prize, awarded by the Norwegian Data
Association for exceptional professional achievements; the John von
Neumann Medal, awarded by the Institute of Electrical and Electronics
Engineers (IEEE); and the A. M. Turing Award, issued by the Association
of Computing Machinery (ACM) for 2001. In August 2000, he was made
Commander of the Order of Saint Olav by the King of Norway.
Bjarne Stroustrup designed and implemented C++. He pioneered
the use of object-oriented and generic programming techniques in applica-
tion areas where efficiency is at a premium—for example, switching sys-
tems, simulation, graphics, embedded systems, and scientific computation.
The C++ Programming Language (Addison-Wesley, 1985, 1991, 1997, and
“special” edition in 2000) has been translated into 14 languages. In addition
to his five books, Stroustrup has published more than 60 academic and
more popular papers.
Dr. Stroustrup (Ph.D. in computer science from Cambridge University) is
an ACM Fellow and has received that society’s Grace Murray Hopper award
for his work in C++. He is an ATT Bell Labs Research Fellow and an ATT
Fellow and currently holds the position of professor at Texas A&M University.

SIMULA
SIMULA did not start as a programming language, and during
all of its development stages, reasoning outside traditional
programming played an important part in its design.

From the very outset SIMULA was regarded as a system

description language.... [emphasis Nygaard’s]

SIMULA 67, however, was a general-purpose programming language. Was

the original purpose abandoned, or did SIMULA have a “dual personality”? The
 Chapter 2 Philosophical Context 39

latter option is correct. This will be important when we examine what Smalltalk
and C++ chose to borrow from SIMULA.
The concept of SIMULA began with an analysis of operations research and
the kind of complex systems being modeled and analyzed in that domain. The
first goal was to develop a “useful and consistent set of concepts” for modeling
the “structure and interaction” of elements in complex systems. The initial
objectives for the language were as follows:

1. The language should be built around a general

mathematical structure with few basic concepts. This structure
should furnish the operations research worker with a
standardized approach in his description so that he can easily
define and describe the various components of the system in
terms of these concepts.

2. It should be unifying, pointing out similarities and

differences among various kinds of network systems.

3. It should be directing, and it should force the operations

research worker to consider all aspects of the network.

4. It should be general and allow the description of very wide

classes of network systems and other systems that can be analyzed
by simulation, and it should for this purpose contain a general
algebraic dynamic language, such as ALGOL or FORTRAN.

5. It should be easy to read and to print, and it should be suitable

for communication among scientists studying networks.

6. It should be problem oriented and not computer oriented,

even if this implies an appreciable increase in the amount of
work that has to be done by the computer.

Perhaps the most dramatic of these objectives was number 6, which is at

odds with typical objectives for programming languages: efficiency, speed of
execution, smallest possible executable footprint, and more intuitive and useful
representations for computer primitives (memory addresses, operation codes,
and so forth). Subsequent statements of objectives for SIMULA significantly
modified this original goal.
One reason for the change was market driven: “The success of SIMULA
would, regardless of our insistence on the importance of problem orientation,
to a large extent depend on its compile and run-time efficiency as a programming
language” (Wexelblat, p. 447). The other reason for the change was a perceived
40 Object Thinking

lack of conflict between problem orientation and computer orientation. SIMULA’s

developers discovered that “good system description capabilities seem to result
in a more simple and logical implementation,” thereby reducing the load on the
computer’s capabilities.
The focus on problem description—and the resulting simplification of the
implementation—promoted by SIMULA’s developers is paralleled in a paper
called “On Decomposition,” by David Parnas. Parnas’s paper examined two
conceptual abstractions for decomposing complex systems for the purposes of
developing software. One was top-down functional decomposition, the
approach that was gaining widespread acceptance under the label “structured
design.” Functional decomposition is based on an attempt to model the perfor-
mance of the computer and software and to translate the requirements of the
domain problem into those computer-based constructs.
Parnas offered an alternative approach called “design decision hiding,” in
which the problem or problem domain is modeled and decomposed without
consideration of how the component parts of that domain or problem would be
implemented. He was able to show that his alternative led to simpler, easier to
read, easier to maintain, and more composable software modules than func-
tional decomposition. Unfortunately, his advice was essentially ignored as the
juggernaut of structured development came to dominate, at least officially, the
manner in which software was conceived and implemented.
Both Parnas and the SIMULA team point to an important principle.
Decomposition into subunits is necessary before we can understand, model,
and build software components. If that decomposition is based on a “natural”
partitioning of the domain, the resultant models and software components will
be significantly simpler to implement and will, almost as a side effect, promote
other objectives such as operational efficiency and communication elegance. If,
instead, decomposition is based on “artificial,” or computer-derived, abstrac-
tions such as memory structures, operations, or functions (as a package of oper-
ations), the opposite results will accrue.
Remember the minor reason for selecting a programming language noted
earlier: only if performance mandates cannot be satisfied with effective design
is it appropriate to consider selecting a language based on whether it provides
more direct access to and control of hardware. Nygaard’s and Parnas’s observa-
tions are consistent with that assertion. Object thinking leads to better designs
that reduce the demand placed on the machine, so raw efficiency and speed are
far less critical than presumed by most developers.
As the SIMULA team expanded their understanding of the simulation
problem domain and of the kinds of systems to be simulated, they identified
needs for more generalized and interrelated components. These components
had to be implemented, and implementation required adding sophistication to
 Chapter 2 Philosophical Context 41

the language, resulting in the ideas of objects, classes of objects, data and
implementation hiding, virtual procedures, and inheritance.
The legacy of SIMULA is twofold. First and most important from the per-
spective of object thinking, it provided an orientation (a philosophy) of giving
primary importance to understanding and modeling the problem domain. This
philosophy suggested the need for an elegant and powerful language that
would allow direct mapping of components in a domain to the modules
employed in the computer. Second, a number of original concepts, with appro-
priate vocabulary (object, class, inheritance), and some important implementa-
tion tricks (such as abstract data types and compiler-generated structures) were
invented or advanced as the language developed. What use did the inheritors
of SIMULA make of this legacy?

C++
Bjarne Stroustrup was motivated by the desire to create “a better C.” The C pro-
gramming language is noted for its power and conformity to machine architec-
ture, which ensures that C programs are maximally efficient in terms of machine
resources. This same power, however, made its misuse almost inevitable. Bugs
were easy to create and difficult to track down. Too many C programmers
lacked the discipline necessary to properly use the language. In SIMULA,
Stroustrup saw a model for introducing discipline into the C language:

C++ was designed to provide SIMULA’s facilities for program

organization together with C’s efficiency and flexibility for
systems programming.... While a modest amount of innovation
did emerge over the years, efficiency and flexibility have been
maintained without compromise.

Stroustrup was concerned with creating a “suitable tool” for projects such
as the writing of “a significant simulator, an operating system, and similar
systems programming tasks.” His focus was on the machine—system-level
programming—and on the program. Even though he found SIMULA to be an
excellent tool for describing systems and directly mapping application concepts
into language constructs, he seemed to be more concerned with performance
features of SIMULA than its descriptive capabilities.

SIMULA’s class-based type system was a huge plus, but its

run-time performance was hopeless:

The poor runtime characteristics were a function of the

language and its implementation.... The overhead problems
42 Object Thinking

were fundamental to SIMULA and could not be remedied.

The cost arose from several language features and their
interactions: run-time type checking, guaranteed initialization
of variables, concurrency support, and garbage collection....

SIMULA was conceived to make it easier to describe natural systems and

simulate them in software, even if that meant the computer had to do more
work. The inefficiencies noted by Stroustrup were indeed intrinsic to the lan-
guage and the paradigm created by SIMULA. Stroustrup essentially rejected the
SIMULA philosophy because his problem domain was the computer itself and
performance was the primary goal:

C with Classes [precursor to C++] was explicitly designed to

allow better organization of programs; “computation” was
considered a problem solved by C. I was very concerned that
improved program structure was not achieved at the expense
of run-time overhead compared to C. The explicit aim was to
match C in terms of run time, code compactness, and data
compactness. To wit: someone once demonstrated a three
percent systematic decrease in overall run-time efficiency
compared with C. This was considered unacceptable and the
overhead was promptly removed.

C++ maintained the goal of adding program structure without a loss of

performance. The constant measure of the language was the machine, either
the physical computer platform or the virtual machine—the program. This con-
stant focus on performance limited what could be borrowed from SIMULA—
most important, SIMULA’s goal of being a general systems description language.
Although the claim is made that C++ is a general-purpose programming
language, that assertion should be modified. C++ is a general-purpose language
for describing and efficiently implementing programs that model software
implementation constructs (for example, control structures, data structures, and
algorithms), virtual machines, or hardware elements.
The focus on the machine is C++’s greatest strength and its greatest weak-
ness. Stroustrup explicitly rejected the philosophy and values behind SIMULA
and merely borrowed some of its implementation tricks, thereby creating a lan-
guage that inhibits the direct expression of application designs in any domain
except that of the computer itself. The “simple and logical implementations”
observed by Nygaard and advocated by Parnas of noncomputer problem
domain designs cannot be expressed in C++ without some degree of compro-
mise with those principles upon which the language is predicated.
 Chapter 2 Philosophical Context 43

Smalltalk
Philosophically, Smalltalk’s objects have much in common with
the monads of Leibniz and the notions of 20th century physics
and biology. Its way of making objects is quite Platonic in that
some of them act as idealizations of concepts—Ideas—from
which manifestations can be created. That the Ideas are
themselves manifestations (of the Idea-Idea) and that the Idea-
Idea is a-kind-of Manifestation-Idea—which is a kind-of-itself,
so that the system is completely self-describing—would have
been appreciated by Plato as an extremely practical joke.

Alan Kay describes Smalltalk as a “crystallization of style” language, one

that is an expression of “the insight that everything we can describe can be rep-
resented by a single kind of behavioral building block”—in essence, an object.
From the outset, Kay is characterizing Smalltalk as deriving from the same goals
that motivated SIMULA—a desire to have a simple and expressive language for
describing and representing (simulating) naturally occurring complex systems.

Object-oriented design is a successful attempt to qualitatively

improve the efficiency of modeling the ever more complex
dynamic systems and user relationships made possible by the
silicon explosion.

Note the absence of any reference to computer or program efficiency or

organization. When Kay first encountered SIMULA and its objects and object
manipulation constructs, he experienced a kind of epiphany, as ideas from
mathematics, philosophy, and biology came together.

Bob Barton had said...“The basic principle of recursive design

is to make the parts have the same power as the whole.” For the
first time I thought of the whole as the entire computer and
wondered why anyone would want to divide it up into weaker
things called data structures and procedures. Why not divide
it up into little computers. ... I recalled the monads of Leibniz,
the “dividing nature at its joints” discourse of Plato, and
other attempts to parse complexity. ... It is not too much of an
exaggeration to say that most of my ideas from then on took
their roots from SIMULA—but not as an attempt to improve it.
It was the promise of an entirely new way to structure
computations that struck my fancy.
44 Object Thinking

Another major stream of influences that shaped the development of Small-

talk was education and cognitive theories—ideas about how people (often chil-
dren) think or can be encouraged to develop thinking skills. The computer, for
Kay, promised a potential vehicle for supporting and promoting thinking, the
foundation for an alternative advocated by Marvin Minsky.

It was clear that education and learning had to be rethought

in the light of 20th century cognitive psychology and how good
thinkers really think. Computing enters as a new representation
system with new and useful metaphors for dealing with
complexity, especially of systems.

Although Kay’s account of the origins of Smalltalk addresses issues of

machine efficiency (almost always in the context of making performance con-
form to human user expectations for dialog) and compactness (an overriding
goal was to create a Dynabook or at least a notebook computer), the goals of
machine efficiency and compactness are overwhelmed by other goals. Expres-
siveness in describing complex systems, support for education and dialogic
interaction between children and machines, and even a search for beauty in
programming languages are important examples.

One part of the perceived beauty of mathematics has to do

with a wondrous synergy between parsimony, generality,
enlightenment, and finesse. ... When we turn to the various
languages for specifying computations, we find many to be
general and few to be parsimonious. For example, we can
define universal machine languages in just a few instructions
that can specify anything that can be computed. But most of
those we would not call beautiful, in part because the amount
and kind of code that has to be written to do anything
interesting is so contrived and turgid. ...

A fertilized egg that can transform itself into the myriad of

specializations needed to make sure a complex organism
has parsimony, generality, enlightenment, and finesse—in
short, beauty. ... Nature is wonderful at both elegance and
practicality—the cell membrane is partly there to allow
evolutionary kludges to do their necessary work and still be
able to act as components by presenting a uniform interface
to the world.
 Chapter 2 Philosophical Context 45

Alan Kay clearly was not interested so much in what went on inside the
machine as in how the existence of the machine redefined the act of commu-
nication between person and machine. He saw the personal computer as a
potentially liberating and creative device, but its potential impact was inhibited
by the mode of communication. He saw that a better language, not program-
ming language, was required.
We should derive two lessons and one assertion from this brief historical
retrospective:

■ Lesson The true differences among programming languages are

those that reflect philosophical ideals and values. Those values and
ideals, in turn, determine the degree to which a language naturally
and simply expresses design concepts without resorting to “con-
trived and turgid” code.
■ Lesson If you think about design using an implementation
language—as programmers and especially extreme programmers are
wont to do—your designing will be enhanced or severely restricted
by that language.
■ Assertion If your development project involves modeling, design-
ing, and solving problems in the domain defined by the boundaries of
the computer itself (for example, operating systems, device drivers,
network infrastructure), you will best be served by languages such as
C++, C#, and Java. If, like the vast majority of software developers, you
are interested in modeling, designing, and solving problems in an
application space, you will be far better served by languages such as
Smalltalk, Lisp, FORTRAN, COBOL, and Visual Basic (with Smalltalk
and Visual Basic being generally applicable to a wider variety of appli-
cation domains).

Of course, choice of language is not the only implementation decision that

affects design in a particular way. The decision to employ a relational database
is an implementation decision with the same kind of implications as selecting a
language. Consider the following example. A customer in the real world might
use many different addresses for different reasons. A domain-reflective object
model might have the partial diagrammatic representation in Figure 2-1.
Address is a collection—a recurring field—and therefore, according to the dic-
tates of relational database design, cannot be an attribute of Customer. In fact,
we have to create two entities with a relationship between them, as shown in
Figure 2-2.
46 Object Thinking

Customer

custId
description
addresses
creditData

Object modeling allows

‘attributes’ with
multiple values, e.g.,
description or address.

Figure 2-1 Domain-reflective customer model (partial).

F02MQ01

Customer Address CreditCustomer Normalization requires distribution

of information about a customer
custId addressId cCustId across several entities.
height custId custId
weight street creditLimit
eyecolor street2
city
state
zip

Figure 2-2 Relational customer model (partial).

F02MQ02

In a store I might ask a customer, “Where would you like your purchase
delivered?” The customer would think a bit and give me back an address. But
in the relational example, I cannot ask the customer this question because the
customer does not know the answer. Instead, I have to ask the customer for his
customer number and then ask the collection of addresses in the Address rela-
tion, “Which of you belongs to customer custNo and also has the type value
delivery?” The implementation code gets even more “contrived and turgid”
when I attempt to account for the fact that the customer might want to use dif-
ferent delivery addresses at different times or for different situations. Analogous
design problems occur when you use strongly typed languages—the real world
is pretty fuzzy when it comes to classification—or other constructs that effec-
tively represent the computer but not the application domain.
Programming languages are concrete manifestations of a set of values, ideals,
and goals, which in turn reflect a more inclusive worldview or philosophical
context. Sometimes the language designer makes explicit philosophic deci-
sions, as did the three language developers just discussed. At other times, the
philosophy that shapes the programming language is invisible even to the
designer. Designers are humans and participate in a human culture. Most
humans are oblivious to their own culture: as the saying goes, “as blind to their
own culture as a fish is to the water in which it swims.” It would be expected,
therefore, that the general culture in which language designers live and work
 Chapter 2 Philosophical Context 47

also plays a role in shaping the design of the language. And programming
languages are not the only artifact shaped by a prevailing culture. So too are
methods, processes, and even models.

Note The philosophy behind Smalltalk (and object thinking) does not
make the problem of selecting a correct address trivial, nor does it solve
problems with classification, but it does lead you in the direction of
simpler-to-implement and easier-to-modify solutions. For example, we
ask a human customer, “Where would you like this shipped?” Object
thinking suggests we create a customer object that can respond to
essentially the same message: shipTo. We assume that the customer
has a collection of possible addresses in mind and a rule that they
employ to decide which address to use. We probably assume that the
rule is context sensitive in some way, evaluating to different addresses
at different times. If we simulate all three of these objects—Customer,
addressCollection, and selectionRule —and distribute the problem of
“which address?” across the three, the solution is far simpler than what
would be necessary using a relational schema and an SQL query. The
classification problem is finessed, in a way, rather than solved by object
thinking and Smalltalk philosophy, which suggests that the real world is
fuzzy and any kind of hard-and-fast classification should be avoided
unless absolutely necessary. (Protecting programmers from making
errors is not an absolute necessity.) And when making classifications,
use set criteria that are intrinsic (such as DNA) instead of extrinsic (such
as presence of hair, warm blood, and lays eggs).

We now return to an issue raised at the very beginning of this chapter:

why is XP perceived as a radical departure from accepted practice, and why
does it evoke such strong reactions—both positive and negative—from soft-
ware development practitioners and theorists? The answer to this question lies
in the philosophical context within which software development emerged.
Given that most of computing and software development emerged in the
Western world,2 a common worldview or philosophical context might be assumed.

2. This is a historical observation, not an expression of ethnocentrism or an attempt to claim computing for
Europe and the United States. As will be seen in the rest of the discussion, this is more of an indictment
of the history of computing than a boast. Clearly the roots of computing and many of the most important
contributions to our understanding of computing come from many different places and cultures. The
argument will be made in the next few pages that most of computing, however, is firmly grounded in a
philosophical tradition that arose in Europe and formed the foundation of the Age of Reason.
48 Object Thinking

If this were true, all programming languages, methods, approaches, tools, and pro-
cesses would share a common, albeit deeply hidden, philosophical foundation.
This is clearly not the case. Every culture contains variations, or subcultures,
which exist at various levels of scale. In the next section, we’ll explore two com-
peting philosophical themes that have long been present in Western culture.

Formalism and Hermeneutics

For philosophers, the eighteenth century is considered the Age of Enlighten-
ment, or the Age of Reason. Other descriptive labels for the era include the Age
of Science and the beginning of the Age of the Machine. The universe was con-
sidered a kind of complicated mechanism that operated according to discover-
able laws. Once understood, those laws could be manipulated to bend the
world to human will. Physicists, chemists, and engineers daily demonstrated the
validity of this worldview with consistently more clever and powerful devices.

Forward Thinking
Metaphor and Initial Stories
In this particular case, our system metaphor is easy: it’s the vending machine
itself since our task is to provide software simulations of the behaviors
expected of a Universal Vending Machine. Even more helpful, we have an
entire row of vending machines in the hallway just outside our office that
we can use as references as we talk about what we are building.
The team sits down with the on-site customer to discuss the project.
We have all heard the CEO’s excited description of the UVM, so we have
a common starting point. The first cards that the customer writes include
the following. (The story title precedes the colon; the story narrative
follows the colon.)

■ Payment: The UVM accepts payment.

■ Selection: The UVM allows the customer to make a selection.
■ Change: The UVM dispenses change.
■ Inventory: The UVM updates its inventory.
■ Dispense: The UVM dispenses the selected product.

Rather obvious stories, but a starting point. The team discusses

whether it can estimate the stories provided, decides that it cannot,
and works with the on-site customer to refactor the story list. This effort
yields a modified list:
 Chapter 2 Philosophical Context 49

Forward Thinking (continued)

1. VerifyCC: The UVM verifies the credit card.

2. VerifyDBT: The UVM verifies the debit card.
3. ChargeCC: The UVM charges the credit card.
4. DeductFunds: The UVM deducts funds from the debit card.
5. Cash: The UVM accepts cash.
6. CurrencyConversion: The UVM converts currencies.
7. WebConnect: The UVM accepts a Web connection.
8. Menu: The UVM informs the customer of what is available to pur-
chase, making sure only in-stock items are presented.
9. Selection: The customer is allowed to make a selection. Do we allow
her to change her mind?
10. ValidateSelection: Make sure that the amount of money available is
equal to or greater than the cost of the selected item.
11. ChargeIt: Charge the credit card.
12. DeductIt: Deduct funds from the debit card.
13. DispenseItem: Give the customer what she asked for.
14. Change?: Calculate whether change is due.
15. DispenseChange: Return change to customer.
16. NoChange: Ask the customer to enter correct change only, or offer
to post the change amount to a credit card or debit card instead of
dispensing money.

The next step is to turn the stories into objects. Yes, yes—the next
explicit XP step is to write tests, but what tests, and what should the focus
of the tests be? You know that the code you produce will consist of class
definitions and methods. You know that the refactoring you will eventu-
ally do involves breaking up methods and redistributing methods among
your classes. So even though your next activity may be to write tests, your
next thinking will involve finding objects.
Because you have internalized object thinking, you can look at the
list of stories and foreshadow some likely outcomes. Stories 1 and 2 are
likely to involve the same set of objects and might be a single story. The
same thing is true of stories 13 and 15—dispensing is dispensing.

(continued)
50 Object Thinking

Forward Thinking (continued)

Developers separate into pairs, select a story, and begin develop-

ment. Everyone is confident that we can begin development with such
sparse definition of requirements and in the absence of any kind of formal
modeling. Everyone fully expects a final solution to emerge from ongoing
discussions, elaborations, and implementations of this initial set of stories.
No one feels the need to create syntactically precise models, but it is sur-
prising how much gets jotted down and drawn in the process of discuss-
ing and writing code. Interestingly enough, those that have been doing
objects longer tend to make fewer notes as they proceed, while those that
are relatively new to objects have many more bits of paper on and near
their workspace. Object thinking is a mental exercise: all the ideas, defini-
tions, heuristics, and models presented in this book are internalized as
developers put them into practice.

Writers such as Descartes, Hobbes, and Leibniz provided the philosophical

ground that explained the success of science and extended the mechanical met-
aphor to include human thought. For these thinkers, the universe comprised a
set of basic elements: literally, the periodic table of elements in the case of chem-
istry, properties such as mass for the physicists, and mental tokens in the case of
human thought. These basic elements could be combined and transformed
according to some finite set of unambiguous rules: the laws of nature in the case
of the physical sciences, or classical logic in the case of human thought.

Note Descartes, Hobbes, and Leibniz are likely to be familiar to most

readers as rationalist philosophers. Perhaps less well known is the fact that
all three were convinced that human thought could be simulated by a
machine, a foundation idea behind classical artificial intelligence (AI)
research in the 1970s. All three built or attempted to build mechanical
thinking/calculating devices. Leibniz was so entranced with binary arith-
metic (he invented aspects of binary logic) that it influenced his theology—
“The void is zero and God is one and from the One all things are
derived”—and prompted an intense interest in the Chinese I Ching (Book
of Changes), which has a binary foundation.
 Chapter 2 Philosophical Context 51

This tradition of thought, this worldview or paradigm, has been labeled for-
malism. Other names with various nuances of meaning include rationalism,
determinism, and mechanism. Central to this paradigm are notions of centralized
control, hierarchy, predictability, and provability (as in math or logic). If someone
could discover the tokens and the manipulation rules that governed the universe,
you could specify a syntax that would capture all possible semantics. You could
even build a machine capable of human thought by embodying that syntax in its
construction.
As science continued to advance, other philosophers and theoreticians
refined the formalist tradition. Russell and Whitehead are stellar examples. In
the world of computing, Babbage, Turing, and von Neumann ensured that
computer science evolved in conformance with formalist worldviews. In some
ways, the ultimate example of formalism in computer science is classical artifi-
cial intelligence, as seen in the work of Newell, Simon, and Minsky.
Formalist philosophy has shaped Western industrial culture so extensively
that even cultural values reflect that philosophy. For example, scientific is good,
rational is good, and being objective is good. In both metaphor (“Our team is
functioning like a well-oiled machine”) and ideals (scientific management,
computer science, software engineering), Western culture-at-large ubiquitously
expresses the value system derived from formalist philosophy.
Computer science is clearly a formalist endeavor. Its roots are mathematics
and electrical engineering. Its foundation concepts include data (the tokens), data
structures (combination rules), and algorithms (transformation and manipulation
rules). Behind everything else is the foundation of discrete math and predicate
calculus. Structured programming, structured analysis and design, information
modeling, and relational database theory are all prime examples of formalist think-
ing. These are the things that we teach in every computer science curriculum.
As a formalist, the computer scientist expects order and logic. The “goodness”
of a program is directly proportional to the degree to which it can be formally
described and formally manipulated. Proof—as in mathematical or logical proof—
of correctness for a piece of software is an ultimate objective. All that is bad in soft-
ware arises from deviations from formal descriptions that use precisely defined
tokens and syntactic rules. Art has no place in a program. In fact, many formalists
would take the extreme position: there is no such thing as art; art is nothing more
than a formalism that has yet to be discovered and explicated.
Countering the juggernaut of formalism is a minority worldview of equal
historical standing, even though it does not share equal awareness or popularity.
Variously known as hermeneutics, constructivism, interpretationalism, and
most recently postmodernism, this tradition has consistently challenged almost
everything advanced by the formalists. Iterative development practices, includ-
ing XP, and object thinking are consistent with the hermeneutic worldview.
52 Object Thinking

Unfortunately, most object, XP, and agile practitioners are unaware of this
tradition and its potential for providing philosophical support and justification
for their approach to software development.
Hermeneutics, strictly speaking, is the study of interpretation, originally
the interpretation of texts. The term is used in religious studies in which the
meaning of sacred texts, written in archaic languages and linguistic forms, must
be interpreted to a contemporary audience. Husserl, Heidegger, Gadamer,
Dilthey, and Vygotsky are among the best-known advocates of hermeneutic
philosophy.

Note Hermeneutics (her-me-NOO-tiks or her-me-NYOO-tiks) is derived

from the name of the Greek god Hermes, the messenger or god of
communication. It is a difficult name and does not flow easily off the
tongue like “formalism.” Unfortunately, there is no comfortable alternative
term to use. Most of the philosophers most closely associated with this
school of thought—excepting Heidegger—are probably unknown to
most readers. Unfortunately, there isn’t space to fully explicate the
ideas of these individuals in this book. It’s strongly suggested, how-
ever, that your education—and your education as a software devel-
oper in particular—will not be complete without a reasonably thorough
understanding of their ideas. The bibliography contains a section of
references that can help the reader begin exploring hermeneutics.

The ideas of the hermeneutic philosophers are frequently illustrated with

examples from linguistics, but hermeneutic principles are not limited to that
domain. For example, words (the tokens of thought, according to formalists) do
not have clear and unambiguous meaning. The meaning (semantics) of a word
is negotiated, determined by those using it at the time of its use. Semantics are
ephemeral, emerge from the process of communication, and are partially
embodied in the minds of those involved in creating them.
According to the hermeneutic position, the meaning of a document—say
a Unified Modeling Language (UML) class diagram—has semantic meaning only
to those involved in its creation. However precise and correct the syntax of
such a diagram may be, a significant portion of its meaning (its semantics) is not
in the diagram but exists only in the minds of the developers that created the
diagram. If a team of analysts and designers create a UML diagram and pass it
to a team of programmers to implement, the programmers will not be able to
 Chapter 2 Philosophical Context 53

discern the “meaning” of the diagram. The programmers will, of necessity, have
to interpret the document and find their own meaning (semantics) in its syntax.
And of course, it will be the programmer’s semantics, not the analyst’s/
designer’s syntax, that actually get implemented.
The hermeneutic conception of the natural world claims a fundamental
nondeterminism. Hermeneuticists assert that the world is more usefully
thought of as self-organizing, adaptive, and evolutionary with emergent prop-
erties. Our understanding of the world, and hence the nature of systems we
build to interact with that world, is characterized by multiple perspectives and
constantly changing interpretation. Contemporary exemplars of this paradigm
are Gell-Mann, Kauffman, Langton, Holland, Prigogine, Wolfram, Maturana,
and Varela.

Note Murray Gell-Mann, Stuart Kauffman, Christopher Langton, and

John Holland are closely associated with the Santa Fe Institute and
the study of complexity and artificial life. This new discipline chal-
lenges many of the formalist assumptions behind classical science,
suggesting that significant portions of the real world must be under-
stood by using an alternative paradigm based on self-organization,
emergent properties, and nondeterminism. Ilya Prigogine is a Nobel
prize–winning physicist (as is Gell-Mann) whose work laid many of the
foundations for the study of emergent and chaotic physical systems.
Steven Wolfram, developer of Mathematica and an expert in cellular
automata, has recently published A New Kind of Science (Wolfram
Media, Inc., 2002), which suggests that all we know can be best
explained in terms of cellular automata and emergent systems.
Humberto Maturana and Francisco Varela are proponents of a “new
biology” consistent with complex systems theory and are collaborators
with Terry Winograd on a hermeneutic theory of design strongly influ-
enced by the philosophy of Heidegger.

As exotic and peripheral as these ideas may seem, they have been at the
heart of several debates in the field of computer science. One of the best exam-
ples is found in the area of artificial intelligence—the formalists, represented by
Newell and Simon, arguing with the Dreyfus brothers and others representing
hermeneutic positions.
54 Object Thinking

Note Allen Newell (before his death) and Herbert Simon were
among the leading advocates of traditional artificial intelligence—the
theory that both humans and machines are instances of “physical
symbol systems.” According to them, both humans and machines
“think” by manipulating tokens in a formal way (Descartes redux), and
therefore it is perfectly possible for a digital computer to “think” as well
as (actually better than) a human being. Hubert L. Dreyfus, working
with his brother, Stuart, has been one of the most vocal and visible
critics of traditional AI. What Computers Can’t Do (HarperCollins,
1979) and What Computers Still Can’t Do (MIT Press, 1992), written
by Dreyfus, present arguments based on the work of Husserl and
Heidegger against the formalist understanding of cognition.

Another example centers on the claim for emergent properties in neural net-
works. Emergence is a hermeneutic concept inconsistent with the formalist idea
of a rule-governed world. Arguments about emergence have been heated. The
stronger the claim for emergence by neural network advocates, the greater the
opposition from formalists. Current work in cellular automata, genetic algorithms,
neural networks, and complexity theory clearly reflect hermeneutic ideas.3

Note Marvin Minsky is another leading advocate of traditional AI.

Originally, he was vehemently against the idea of emergent properties
in systems—a view that has seemed to soften in later years. His book
Society of Mind (Touchstone Books, 1988) attempted to use object-
oriented programming ideas to develop a theory of cognition that
could rely on interactions of highly modularized components without
the need for emergent phenomena.

The hermeneutic philosopher sees a world that is unpredictable, biological,

and emergent rather than mechanical and deterministic. Mathematics and logic
do not capture some human-independent truth about the world. Instead, they
reflect the particularistic worldview of a specific group of human proponents.

3. Do not confuse formalism with the use of formal tools, such as mathematics, that are employed in the
cited fields of study.
 Chapter 2 Philosophical Context 55

Software development is neither a scientific nor an engineering task. It is an act

of reality construction that is political and artistic.
As you can see, formalism and hermeneutics contest each other’s basic
premises and assumptions about the nature of the universe and the place of
humanity within that universe. Challenges to basic assumptions are frequently
challenges to core values as well. And often, fundamental assumptions and core
values are seldom examined. Like articles of faith, they are blindly defended.
As noted previously, Western culture in general is largely formalist (using
the labels rationalist and scientific rather than formalist) in its orientation. Any-
thing challenging this position is viewed with suspicion and antagonism. It is
for this reason that the conflict between hermeneutic and formalist worldviews
frames the debate about an object paradigm.

Forward Thinking
An Observation
A quick glance around the development room reveals some interesting
graphical models drawn on paper or on the whiteboard. We also see peo-
ple pointing to and modifying these models as they engage in developing
tests or writing code. Figure 2-3 and Figure 2-4 are examples of two such
artifacts. Figure 2-3 is just a rectangle with some text, but it seems to be a
model of an object that one team is working on. Figure 2-4 consists of
some labeled boxes, lines, and arrows; judging from the conversation of
the development pair, it’s a model of object interactions that they are try-
ing to understand as a basis for writing a test.

Figure 2-3 A graphical representation of one aspect of an object: its

F02MQ03

behaviors or responsibilities.

Both of these models are sketches—nothing formal about them.

Both are tools to facilitate communication among the developers. Neither
tool is likely to be useful to those outside the team. Neither contains any
“truth.” Both are ways for the developers to explore and share the thinking

(continued)
56 Object Thinking

Forward Thinking (continued)

in their individual heads with each other. Both provide a kind of external
memory for those involved in the development activities at hand. Neither
is worth keeping around once the task that prompted their creation is
completed.

Figure 2-4 A graphical model of objects communicating (interacting)

F02MQ04

with one another.

These artifacts are quite consistent with hermeneutic philosophy

and at odds with formalist philosophy. They are quite useful for extreme
programmers.

XP is the most recent example of a series of attempts to apply hermeneutic,

human-centric, and aformal ideas to software development. Some antecedents
include the “Greek versus Roman” software development cultures identified by
Robert L. Glass4 and the associated debates over the role of creativity in software

4. Glass, Robert L. Software Creativity. Englewood Cliffs, NJ: Prentice Hall, 1995.
 Chapter 2 Philosophical Context 57

development, the conflicts between “fuzzies” and “neats” in AI, and the classic
debates between devotees of Smalltalk and C++.

Behind the Quotes

Robert L. Glass
A prolific writer and chronicler of ideas in software development as well as
a leading practitioner, Robert L. Glass appears frequently in publications
ranging from ACM Communications to his own newsletter, The Software
Practitioner. His extensive experience in the real world and the world of
academe make his insights into the conflict between how it is done and
how theorists think it is done invaluable for everyone involved in any kind
of software development. One of his main themes is that practice requires a
great deal more creativity and experimentalism than computer scientists,
software engineers, and academicians are willing to publicly acknowledge.
His latest work, Facts and Fallacies of Software Engineering (Addison-
Wesley, 2003), presents in a concise and highly readable fashion many of the
critiques of formalist software development methodology that are presented
here as foundation positions for object thinking and extreme programming.

More recently, Michael McCormick notes that

What XP uncovered (again) is an ancient, sociological San

Andreas Fault that runs under the software community—
programming versus software engineering (a.k.a. the scruffy
hackers versus the tweedy computer scientists). XP is only the
latest eruption between opposing continents.5

XP is the latest assertion of the view that people matter. XP is the latest chal-
lenger to the dominant (and hostile) computing and software engineering cul-
ture. XP is the latest attempt to assert that developers can do the highest-quality
work using purely aformal methods and tools. And XP is the latest victim of the
opprobrium of the formalists and the latest approach to be described as suitable
only for dealing with small, noncritical problems.
To the extent that objects (our present focus) are seen as an expression of
a hermeneutic point of view, they have been characterized as antirationalist, or

5. McCormick, Michael. “Programming Extremism.” Communications of the ACM 44(6), June 2001, 109–110.
58 Object Thinking

at best nonrationalist, challenges to the prevailing philosophy. It is my assertion

that objects are, or are perceived to be, a reflection of hermeneutic philosophy.
A way to test this assertion is to compare objects with other ideas about soft-
ware development that are clearly hermeneutic—for example, postmodernism.

Postmodern Critiques
There is a controversy smoldering in the computer science
world at the intersection of two important topics: formal
methods and heuristics. The controversy, though it may
sound esoteric and theoretic, is actually at the heart of our
understanding of the future practice of software engineering.

What is meant by formal methods? Techniques, based on a

mathematical foundation, which provide for systematic
approaches to problem solution. What is meant by heuristics?
Techniques that involve trial-and-error approaches to problem
solution.

It should be noted that careful reading will disclose a potential

middle ground. Formal methods, perhaps, are appropriate for
solving mechanistic and well-understood problems or parts of
problems; heuristics are necessary for more complicated and
creative ones.
—Robert L. Glass6

Robert L. Glass probably would not identify himself as a postmodernist,

but he does provide a good transition to a consideration of postmodern philos-
ophy and computer science. The preceding quotes are consistent with the con-
trast between formalist and hermeneutic philosophy discussed in the preceding
section. The third quote additionally suggests the superiority of heuristics over
formalism when we need to address large-scale, complicated (and possibly
complex) systems that involve human beings.
Glass’s arguments in favor of heuristics and creativity in software design
mirror the hermeneutic arguments against formalism. Formal approaches simply
will not work beyond a certain scale. Formalism works close to the computer, is
highly questionable at the level of an application, and fails at the level of com-
plete systems and architectures.

6. Glass, Robert L. Software Creativity. Englewood Cliffs, NJ: Prentice-Hall PTR. 1995. (p. 42)
 Chapter 2 Philosophical Context 59

The work of Terry Winograd provides a complementary parallel to Glass

while making a more direct link to hermeneutic philosophy. Early in his career,
Winograd was a strong advocate of classical AI and was a strong formalist.
Exposure to the ideas of Humberto Maturana and Francisco Varela, along with
the philosophical works of Martin Heidegger, prompted his reconsideration of
AI’s formalist tenets. Winograd has turned his attention from formal modeling
of the world to the issue of designing software to be used in the real world. He
characterizes design as a conscious act, human-centric, conversational or dia-
logic in nature (between artifacts such as software and hardware and human
beings who are users of those artifacts), creative, communicational, with social
consequences, and done as a social activity.
One of the influences on Winograd’s thinking was the work of Humberto
Maturana and Francisco Varela regarding the evolution of autopoietic (self-orga-
nizing) biological and cognitive systems. The structural-coupling mechanism
used by cells to establish cooperative complex structures, and eventually all the
familiar forms of flora and fauna, provided one bridge to the social nature of
computer systems and their development. Another influence was the work of
Martin Heidegger, a hermeneutic philosopher. Writing with Fernando Flores,
Winograd explored the work of Heidegger and its implications for computer
system design.
One of the most important implications was the denial of “intrinsic truth or
meaning” in any artifact—whether it was a computer, a piece of software, or a
simple statement in a natural language. This claim is also central to the school
of thought that has been labeled postmodern. It is also one of the core claims
of all the hermeneutic philosophers. Because of this implication, the design of
computer systems must, for Winograd and other postmodernists, be refocused
on the use of software and hardware as communication devices for a particular
group of people at a particular point in time.
While Winograd and Flores concentrated on Heidegger’s notion of break-
down and issues of communication, Christiane Floyd and her coauthors
extended the discussion to include other facets of postmodern philosophy.7
The role of politics and power relationships in both the imposition of a software
artifact on a community of users and the group dynamics of those charged with
the creation of the software artifact in the first place are central concerns of
Floyd and her colleagues. In a similar vein, Richard Coyne8 addresses the issue
of design in a postmodern age.
The importance of all this work—beginning with Glass’s concerns about
creativity and including the postmodernists’ concerns with computer system

7. Floyd, C., H. Zullighoven, R. Budde, and R. Keil-Slawik (eds.). Software Development and Reality
Construction. Springer-Verlag, 1992. Also Dittrich, Yvonne, Christiane Floyd, and Ralf Klischewski.
Social Thinking, Software Practice. MIT Press, 2002.
8. Coyne, Richard. Software Development in a Postmodern Age.
60 Object Thinking

design—is its clear extension of hermeneutic arguments against formalism. It’s

also the context in which objects as a software development metaphor were
coined. The first expression of “object thinking” by Alan Kay and the research-
ers at Xerox PARC was concerned, like the postmodernists, far more with
people and communication issues than it was with technical computer and
formalist issues.
Knowingly or not, the object community was concerned in the 1960s and
1970s with the same issues raised by the hermeneuticists in the nineteenth
and twentieth centuries, as well as the issues that would be raised by the post-
modernists of the 1990s. To an object advocate, objects are valuable because
they facilitate user/computer interaction and communication among members
of development teams. Objects enhance the art of software development but
not necessarily the engineering.
Adherents to formalist ideas, including many computer scientists and soft-
ware engineers, dismissed objects as irrelevant. When object technology looked
as if it might make serious inroads into real-world development, the formalists
attacked in the same manner that they attack other critiques of formalist
approaches—for example, the creativity discussed by Robert Glass.
Simultaneous with the attack on basic philosophy, the traditionalists began
to lay claim to the form of objects by equating them with the black box module.
They were also quick to adopt less threatening innovations such as abstract data
types as logical extensions of traditional software engineering theory. As a conse-
quence, the object technology that became widely adopted was the formalist
recasting of object ideas rather than the “pure” object paradigm itself.
A similar phenomenon is evident in the patterns movement, wherein many
of the core philosophical ideas of Christopher Alexander, the original inspiration
for the attention paid to patterns, are being dismissed or co-opted by traditional
software developers who like the form of patterns but are uncomfortable with the
more esoteric ideas.

Rejecting Mysticism
Although A Pattern Language: Towns, Buildings, Construction
(Oxford University Press, 1977), by Christopher Alexander, was published
before his The Timeless Way of Building (Oxford University Press, 1979),
the latter provides the philosophical foundation for the former. A Pattern
Language is cited by everyone in the patterns movement as an inspiration
for their own efforts, but The Timeless Way is seldom mentioned. Of
 Chapter 2 Philosophical Context 61

Rejecting Mysticism (continued)

course, it is in the latter book that Alexander’s mysticism is most evident.

Consider the following:

A building or town will only be alive to the extent that it is

governed by the Timeless Way. To seek the Timeless Way we
must first know the Quality Without A Name. To reach the
Quality Without A Name we must then build a living pattern
language as a Gate. Once we have built the Gate, we can pass
through it to the practice of the Timeless Way. And yet the
Timeless Way is not complete, and will not fully generate the
Quality Without A Name, until we leave the Gate behind.

The emphasized statements are actually chapter headers from The

Timeless Way of Building. They read far more like the Taoist (and later
Zen) story of the Boy and the Bull, which is an allegory of the process of
obtaining enlightenment.
It’s not surprising that computer scientists and software engineers,
embedded in a formalist culture, find little of value in this aspect of
Alexander’s work. You have to wonder, however, how deeply they
understand Alexander’s ideas about patterns if they dismiss what are
clearly, for Alexander, fundamental philosophical presuppositions. Is it
not possible, even likely, that they are reading into Alexander’s pattern
ideas their own philosophical biases? And if so, how valuable was
Alexander’s contribution after all?
This example exposes a bias of the author. You cannot claim
to understand something—in this case, object thinking and extreme
programming—unless you are able to understand the form, the sub-
stance, and the presuppositions that support the form and the substance.

Deciding to be an object thinker or an extreme programmer is a decision

to set oneself in opposition, in very important ways, to mainstream software
development thought and practice to become a revolutionary.
Revolutions, especially in the arena of software development, seldom suc-
ceed. They are resisted, and attempts are made to marginalize the proponents of
new ideas. Opponents from the mainstream have already suggested that object
thinking, as presented in this book, and XP are “good,” but only in a small niche,
62 Object Thinking

for small teams, and for noncritical software. What the critics fail to realize, how-
ever, is that object thinking and agile thinking are not a means for solving soft-
ware problems; they are a means for creating better people and better teams of
people. Object thinking and XP will produce a culture (see note), not a technique;
they will give rise to better people capable of attacking any kind of problem and
able to develop systems on any level of complication and scale.

Note A culture already exists that is friendly to object thinking and

extreme programming. As Ken Auer points out, there were hundreds
of people employed by Smalltalk vendors and thousands of projects
written in Smalltalk (many still flourishing). Add in all the people in con-
sulting organizations and all the members of all the project teams, and
you have a sizable number of people. Include the growing number of
XP/agile experts. Even if most of them have failed to grasp the fullness
of object philosophy and culture, the society, with a culture of shared
values and worldviews, clearly exists and is not limited to a handful of
researchers in Palo Alto, consultants in the woods of Oregon, and a
few strange academicians in odd corners of the globe.
From Philosophy
to Culture
Philosophy provides roots, determines some values, and affects the design of
some tools, but the fullness of the object thinking difference must be under-
stood in terms of a broader context—as a culture. The cultural perspective sug-
gests the need to look for the shared, socially learned knowledge (norms,
values, worldviews) and patterns of behavior (individual actions and organiza-
tional relationships) that characterize a group of people. The philosophical
positions discussed in the preceding chapter are part of the cultural knowledge
shared by object thinkers as members of an object culture.
Robert Glass used the metaphor of culture to explain differences in how
different groups of people conceive of and develop software and how the
results of their work are evaluated. His contrast of Roman and Greek cultures
directly parallels the contrast made in the preceding chapter between formal-
ist and hermeneutic philosophies. The Greek culture described by Glass is
a close match to the XP culture1 and the object culture we will explore in
this chapter.

1. West, David. “Enculturating Extreme Programmers,” Proceedings XP Universe. Chapel Hill,

N.C., 2001.

63
64 Object Thinking

Greek and Roman

Robert Glass makes the argument for two cultures within the realm of soft-
ware development, two cultures that frequently find themselves in conflict
based on cultural values. He uses an analogy with Greek and Roman culture
to illustrate the differences as follows:

In ancient Greece, an individual would act as his own agent in

his own behalf, or combine with other people to act together as
a team. In a Greek work environment, you bring your tools to
work with you, you do your stuff, and then you pack up your
tools and take them home. You are an individual—an
independent contractor. You are not owned body and mind.
You are merely providing a service for compensation.

In Rome, one’s first duty was to the group, clan, class, or faction
upon which one depended for status. Known as gravitas, this
meant sacrificing oneself for the good of the organization, and
giving up one’s individuality and identifying closely with the
group. In a Roman environment you go to work, the company
hands you your tools, and then it holds you and your mind
hostage until you sever your relationship with the organization.
You are not an individual: you are owned by the organization
body and mind, twenty-four hours a day. There are substantial
rewards for this, however. The organization provides you with
security, money, and power.

Glass is particularly interested in the degree to which the two cultures

support creativity and asserts that the Roman culture is likely to take the cre-
ativity, passion, and magic out of the work of software development. He
further notes that Roman culture will emphasize up-front planning, control,
formal procedures as a means of control, and maximum documentation and
will value logical, analytical thinking above empirical and inductive thinking.
Even a cursory evaluation of XP values and practices reveals their
incompatibility with Roman thinking. When objects were first introduced,
they too reflected a Greek and not a Roman culture. Smalltalk was moti-
vated by a need to empower people, to make interaction with a computer
fun and creative. Exploratory development, a kind of rapid prototyping,
was seen as the proper way to develop new software—as opposed to the
notion of up-front, detailed design and rote implementation favored by
the (Roman) structured development culture.
 Chapter 3 From Philosophy to Culture 65

Our exploration of the object culture begins with a rough enumeration of

some groups likely to be included in this culture. The original proponents
of object ideas (the original SIMULA team, the Smalltalk team), advocates of
behavior-based object methods, and the first XP and agile practitioners are likely
candidates, and of course, any who recognize themselves as members of the
“Greek” culture described by Glass.
Absent a full ethnography, a cursory look at the object culture reveals
some important traits and characteristics. Specifically:

■ A commitment to disciplined informality rather than defined formality

■ Advocacy of a local rather than global focus
■ Production of minimum rather than maximum levels of design and
process documentation
■ Collaborative rather than imperial management style
■ Commitment to design based on coordination and cooperation rather
than control
■ Practitioners of rapid prototyping instead of structured development
■ Valuing the creative over the systematic
■ Driven by internal capabilities instead of conforming to external
procedures

Cultures will usually have an origin myth, various heroes and heroines, and
stories about great deeds and important artifacts, as well as a set of core beliefs
and values. All of these are evident in the object culture and someday will be cap-
tured in a definitive ethnography. It is not my intent to elaborate that culture here,
merely to draw the reader’s attention to the fact that such a culture exists.
Being aware of object culture is valuable for object thinkers in four ways.
First, it provides insight into the dynamics of interaction (or lack of it) between
objectivists and traditionalists. Often the only way to understand the mutual
miscommunications and the emotional antipathy of the two groups comes from
understanding the underlying cultural conflict.
Second, and most important, it reminds the aspiring object thinker that
he or she is engaged in a process of enculturation, a much more involved
endeavor than learning a few new ideas and adopting a couple of alternative
practices. Most of the material in the remainder of this book cannot be fully
understood without relating it to object culture in all its aspects.
Third, it suggests a way to know you have mastered object thinking. When
all of your actions, in familiar and in novel circumstances, reflect “the right thing”
without the intervention of conscious thought, you are an object thinker.
66 Object Thinking

Fourth, it reminds the object thinker that culture is not an individual thing;
it is rooted in community. To change a culture, you must change individuals
and the way that individuals interact and make commitments to one another.
Culture is shared.
Subsequent chapters will deal with object thinking specifics, all of which
are intimately related to a set of first principles, or presuppositions reflective of
the object culture as a whole. The four principles introduced here are frequently
stated, stated in a manner that implies that the value of the principle is obvious.
Just as a member of any culture makes assertions that are indeed obvious—to
any other member of that culture.

Four Presuppositions
To those already part of the object culture, the following statements are obvious
(“they go without saying”) and obviously necessary as prerequisites to object
thinking:

■ Everything is an object.
■ Simulation of a problem domain drives object discovery and definition.
■ Objects must be composable.
■ Distributed cooperation and communication must replace hierarchi-
cal centralized control as an organizational paradigm.

For those just joining the object culture, each of these key points will need
to be developed and explained. For those opposed to the object culture, these
same presuppositions will be major points of contention.

One: Everything is an object.

This assertion has two important aspects. One is essentially a claim to the effect
that the object concept has a kind of primal status—a single criterion against
which everything else is measured. Another way of looking at this claim would
be to think of an object as the equivalent of the quanta from which the universe
is constructed. The implication of this aspect of everything-is-an-object suggests
that any decomposition, however complicated the domain, will result in the
identification of a relatively few kinds of objects and only objects.
There will be nothing “left over” that is not an object. For example:

■ Relationships, which are traditionally conceived as an association

among objects that is modeled and implemented in a different way
than an object—would themselves become just another kind of
object, with their own responsibilities and capabilities.
 Chapter 3 From Philosophy to Culture 67

■ Data traditionally is seen as a kind of “passive something” fundamen-

tally different from “active and animated things” such as procedures.
Something as simple as the character D is an object—not an element
of data—that exhibits behavior just as does any other object. Whatever
manipulations and transformations are required of an object, even a
character, are realized by that object itself instead of some other kind
of thing (a procedure) acting upon that object. The commonsense
notion of data is preserved because some objects have as their pri-
mary, but not exclusive, responsibility the representation to human
observers of some bit of information.
■ Procedures as a separate kind of thing are also subsumed as ordinary
objects. We can think of two kinds of procedure: a script that allows a
group of objects to interact in a prescribed fashion and the “vital force”
that actually animates the object and enables it to exhibit its behaviors.
A script is nothing more than an organized collection of messages, and
both the collection and the message are nothing more than ordinary
objects. The vital force is nothing more than a flow of electrons
through a set of circuits—something that is arguably apart from the
conceptual understanding of an object, just as the soul is deemed to be
different from but essential to the animation of a human being.

Equating, even metaphorically, a procedure to a soul will strike most readers

as a bit absurd, but there is a good reason for the dramatic overstatement. It some-
times takes a shock or an absurdity to provide a mental pause of sufficient length
that a new idea can penetrate old thinking habits. This is especially true when it
comes to thinking about programming, wherein the metaphysical reality of two
distinct things—data and procedures—is so ingrained it is difficult to transcend. So
difficult, in fact, that most of those attempting object development fail to recognize
the degree to which they continue to apply old thinking in new contexts.
Take object programming, for example—using Smalltalk as an example
merely because it claims to be a pure object language. Tutorials from Digitalk’s
Smalltalk manuals illustrate how programmers perpetuate the notion that some
things “do” and others are “done to.”
The code in Listing One—Pascal is a Pascal program to count unique
occurrences of letters in a string entered by a user via a simple dialog box.
(Pascal was designed to teach and enforce that algorithms [active procedures]
plus [passive] data structures = program mode of thinking.)
Listing Two—Naive Smalltalk shows an equivalent Smalltalk program as
it might be written by a novice still steeped in the algorithms plus data structures
mode of programming. Both programs contain examples of explicit control and
overt looping constructs. The Pascal program also has typed variables—an
implicit nod to the need for control over the potential corruption of passive data.
68 Object Thinking

Listing One—Pascal
program frequency;
const
size 80;
var
s: string[size];
i: integer;
c: character;
f: array[1..26] of integer;
k: integer;
begin
writeln('enter line');
readln(s);
for i := 1 to 26 do f[i] := 0;
for i := 1 to size do
begin
c := asLowerCase(s[i]);
if isLetter(c) then
begin
k := ord(c) – ord('a') + 1;
f[k] := f[k] + 1
end
end;
for i := 1 to 26 do
write(f[i], ' ')
end.

There is some evidence of object thinking in Listing Two—mostly conven-

tions or idioms enforced by the syntax of the Smalltalk language—the use of
the Prompter object, the control loops initiated by integer objects receiving
messages, discovery of the size of the string being manipulated by asking the
string for its size, and so forth.

Listing Two—Naive Smalltalk

| s c f k |
f := Array new: 26.
s := Prompter prompt: 'enter line' default: ' '.
1 to: 26 do: [:i | f at: i put: 0].
1 to: s size do: [
:I | c := (s at: i) asLowerCase.
c isLetter ifTrue: [
k := c asciiValue - $a asciiValue + 1.
f at: k put: (f at: k) + 1.
].
].
^ f

A programmer better versed in object thinking (and of course, the class

library included in the Smalltalk programming environment) starts to utilize the
 Chapter 3 From Philosophy to Culture 69

innate abilities of objects, including data objects (the string entered by the user
and character objects), resulting in a program significantly reduced in size and
complexity, as illustrated in Listing Three—Appropriate Smalltalk.

Listing Three—Appropriate Smalltalk

| s f |
s := Prompter prompt: ' enter line ' default: ' '.
f := Bag new.
s do: [ :c | c isLetter ifTrue: [f add: c asLowerCase]].
^ f.

Types, as implied earlier, create a different kind of thing than an object.

Types are similar to classes in one sense, but classes are also objects and types are
not. This distinction is most evident when variables are created. If variables are
typed, they are no longer just a place where an object resides. Typing a variable
is a nonobject way to prevent all but a certain kind of object from taking up resi-
dence in a named location. Many people have advanced arguments in favor of
typing, but none of those arguments directly challenges the everything-is-an-
object premise. The arguments in favor of types are orthogonal to the arguments
in favor of treating everything as an object. (See note.)

Note Programs are written by human beings, and human beings

make mistakes. One response to this truism is to assume that the quan-
tity of mistakes is both high and essentially constant, which mandates
the existence and use of error detection and prevention mechanisms—
such as typing. Mechanisms, such as typing, are always constrictive, so
much so that every typed language I know of allows ways to escape the
confines of strict typing—casting, for example—which reintroduce the
potential for the errors that typing was intended to prevent. An alterna-
tive response, one consistent with the ideas and ideals of object think-
ing, is to reduce the programmer’s proclivity for making errors by
teaching the programmer the precepts of simplicity and testing. Most of
the time, data moves about a program with little chance of error arising
from the wrong kind of data being in the wrong place at the wrong time.
(User input is the obvious major exception.) If your thinking about
objects and object communication reveals the potential for a type error,
you should create a test for such errors and include an explicit check in
your code at that point. (“Element of data occupying variable X, what
class are you an instance of?”) Since everything is an object, your ele-
ment of data is an object quite capable of telling you its class. You can
have all the benefits of typing without the constraints and the complica-
tions arising from escape valves such as casting.
70 Object Thinking

The everything-is-an-object principle applies to the world, the problem

domain, just as it applies to design and programming. David Taylor2 and Ivar
Jacobson3 use objects as an appropriate design element for engineering, or
reengineering, businesses and organizations. (See the sidebar, “David A. Taylor
and Convergent Engineering.”)

David A. Taylor and Convergent Engineering

Traditional modeling of businesses and organizations is flawed according
to Taylor because of the lack of consistency among the set of models uti-
lized. For example, neither a financial model nor a data model captures
the cost of a bit of information, and inconsistency in design philosophy
prevents the two models from collectively revealing such costs—they can-
not be coordinated.
In his book Business Engineering with Object Technology (John
Wiley and Sons, 1995), Taylor suggests creating a single object model
incorporating everything necessary to produce traditional financial, simu-
lation, process, data, and workflow models as views of the unifying object
model. His process for accomplishing this goal is convergent engineering,
and it, in turn, is based on a behavioral, CRC (Class, Responsibility, Col-
laborator) card, approach to object discovery and specification.
In addition to describing how to conceptualize objects and classes,
Taylor describes a process for discovery and specification leading to the
creation of the organizational object model. That model identifies all the
objects in an organization and how they interact—not just the ones that
will eventually be implemented as software. He also provides a frame-
work for business objects that illustrates the power of object thinking in
generating simple but powerful objects. His framework defines four
classes (Business Elements, Organizations, Processes, and Resources),
describes the behaviors of each, how those behaviors contribute to the
generation of the five standard types of business model, how they can be
customized, and how their interoperation can be optimized to reengineer
the organization as a whole.

2. Taylor, David. Business Engineering with Object Technology. John Wiley & Sons, 1995.
3. Jacobson, Ivar. The Object Advantage: Business Process Reengineering with Object Technology. ACM
Press. Reading, MA: Addison-Wesley. 1994.
 Chapter 3 From Philosophy to Culture 71

The programming example shown earlier illustrates one dimension of

treating everything as an object. Applying the everything-is-an-object principle
to the world—finding and specifying objects that are not going to be imple-
mented in program code or software—can be illustrated by considering a
Human object. Objects, as we will discuss in detail later, are defined in terms of
their behaviors. A behavior can be thought of as a service to be provided to
other objects upon request.
What services do humans provide other objects? For many, this is a sur-
prising question because human beings are not “implemented” by developers
and are therefore considered outside the scope of the system. But it is a fair
question and should result in a list of responsibilities similar to the following:

■ Provides information
■ Indicates a decision
■ Provides confirmation
■ Makes a selection

The utility of having a Human object becomes evident in the simplifica-

tion of interface designs. Acknowledging the existence of Human objects
allows the user interface to reflect the needs of software objects for services
from Human objects. This simple change of perspective—arising from applica-
tion of the everything-is-an-object principle—can simplify the design of other
objects typically used in user-interface construction.
Additional implications of the everything-is-an-object premise will be seen
throughout the remainder of this book.

Two: Simulation of a problem domain drives object discovery

and definition.
Decomposition—breaking a large thing up into smaller, more easily understood
things—is necessary before we can solve most of the problems we encounter as
software developers. There are different approaches to decomposition. For
example, find the data and data structures, find the processing steps, and find
the objects. In object thinking, the key to finding the objects is simulation. The
advocacy of simulation for object discovery has four primary roots:

■ The system description language philosophy behind SIMULA, as dis-

cussed in the preceding chapter.
■ Alan Kay’s ideas about user illusions and objects as reflections of
expectations based on an understanding of how objects behave in a
domain.
72 Object Thinking

■ David Parnas’s arguments in favor of a “design decision hiding”

approach to decomposition—partitioning the problem space and not
the solution space as did functional decomposition approaches—as
discussed in the preceding chapter.
■ Christopher Alexander’s4 ideas about design as the resolution of forces
in a problem space and his subsequent work on patterns that underlie
the organization of a problem space and provide insights into good
design. These will be elaborated later in this book in the discussion of
patterns and pattern languages as an aspect of object thinking.

Proper decomposition has been seen as the critical factor in design from very
early times. This quotation from Plato (which you might recall from Chapter 1) is
illustrative.

[First,] perceiving and bringing together under one Idea the

scattered particulars, so that one makes clear the thing which he
wishes to do... [Second,] the separation of the Idea into classes, by
dividing it where the natural joints are, and not trying to break
any part, after the manner of a bad carver... I love these processes
of division and bringing together, and if I think any other man is
able to see things that can naturally be collected into one and
divided into many, him I will follow as if he were a god.

Plato suggests three things: decomposition is hard (and anyone really

good at it deserves adoration), any decomposition that does not lead to the dis-
covery of things that can be recombined—composed—is counterproductive,
and the separation of one thing into two should occur at “natural joints.” By
implication, if you decompose along natural joints—and only if you do so—you
end up with objects that can be recombined into other structures. Also by impli-
cation, the natural joints occur in the domain, and “bad carving” results if you
attempt to use the wrong “knife”—the wrong decomposition criterion.
If you have the right knife and are skilled in its use—know how to think
about objects and about decomposition—you will complete your decomposi-
tion tasks in a manner analogous to that of the Taoist butcher:

The Taoist butcher used but a single knife, without the need to
sharpen it, during his entire career of many years. When asked
how he accomplished this feat, he paused, then answered, “I
simply cut where the meat isn’t.”

4. Alexander, Christopher. Notes on the Synthesis of Form. Harvard University Press, 1970.
 Chapter 3 From Philosophy to Culture 73

According to this traditional story, even meat has natural disjunctions that
can be discerned by the trained eye. Of course, a Taoist butcher is like the Zen
master who can slice a moving fly in half with a judicious and elegant flick of
a long sword. Attaining great skill at decomposition will require training and
good thinking habits. It will also require the correct knife.
Decomposition is accomplished by applying abstraction—the “knife” used
to carve our domain into discrete objects. Abstraction requires selecting and
focusing on a particular aspect of a complex thing. Variations in that aspect are
then used as the criteria for differentiation of one thing from another. Tradi-
tional computer scientists and software engineers have used data (attributes) or
functions (algorithms) to decompose complex domains into modules that could
be combined to create software applications. This parallels Edsger Wybe Dijkstra’s
notion that “a computer program equals data structures plus algorithms.”

Behind the Quotes

Edsger Wybe Dijkstra
Professor Edsger Wybe Dijkstra, a noted pioneer of the science and indus-
try of computing, died in August 2002 at his home in the Netherlands.
Dijkstra was the 1972 recipient of the ACM Turing Award. (Some
consider this award the Nobel Prize for computing.) He was a member of
the Netherlands Royal Academy of Arts and Sciences and a Distinguished
Fellow of the British Computer Society. He received the 1989 ACM SIGCSE
Award for Outstanding Contributions to Computer Science Education. The
C&C Foundation of Japan recognized Dijkstra “for his pioneering contri-
butions to the establishment of the scientific basis for computer software
through creative research in basic software theory, algorithm theory, struc-
tured programming, and semaphores.” He is credited with the idea of
building operating systems as explicitly synchronized sequential processes
and for devising an amazingly efficient shortest-path algorithm. He
designed and coded the first Algol 60 compiler.
Dijkstra is one of the best examples of the formalist position in
computer science. He believed and argued in favor of the position that
mathematical logic must be the basis for sensible computer program
construction. He added the term structured programming to the language
of our profession and led the fight against unconstrained “GO TO” state-
ments in program code.

(continued)
74 Object Thinking

Behind the Quotes (continued)

Some other common computer science concepts and vocabulary

credited to Dijkstra include separation of concerns (which is important to
object thinking), synchronization, deadly embrace, dining philosophers,
weakest precondition, and the guarded command. He introduced the con-
cept of semaphores as a means of coordinating multiprocessing. The
Oxford English Dictionary credits him for introducing the words vector
and stack into the computing context.

The fact that a computer program consists of data and functions does not
mean that the nonsoftware world is so composed. Using either data or function
as our abstraction knife is exactly the imposition of artificial criteria on the real
world—with the predictable result of “bad carving.” The use of neither data nor
function as your decomposition abstraction leads to the discovery of natural
joints. David Parnas pointed this out in his famous paper “On Decomposition.”
Parnas, like Plato, suggests that you should decompose a complex thing along
naturally occurring lines, what Parnas calls “design decisions.”
Both data and function are poor choices for being a decomposition tool.
Parnas provided several reasons for rejecting function. Among them are the
following:

■ Resulting program code would be complicated, far more so than

necessary or desirable.
■ Complex code is difficult to understand and test.
■ Resulting code would be brittle and hard to modify when require-
ments changed.
■ Resulting modules would lack composability—they would not be
reusable outside the context in which they were conceived and
designed.

Parnas’s predictions have consistently been demonstrated as the industry

blithely ignored his advice and used functional decomposition as the primary
tool in program and system design for 30 years (40 if you recognize that most
object development also uses functionality as an implicit decomposition crite-
rion). Using data as the decomposition abstraction leads to a different set of
problems. Primary among these is complexity arising from the explosion in
total data entities required to model a given domain and the immense costs
incurred when the data model requires modification.
 Chapter 3 From Philosophy to Culture 75

Note Some examples of the kind of explosion referred to in the pre-

ceding paragraph are from my own consulting practice. One organiza-
tion designed a customer support system that identified 15 different
customer classes because they were using a data-oriented approach
and had to create new classes when one type of customer did not
share attributes of the other types. In a much larger example, a com-
pany had just completed a corporate data model (costing millions of
dollars) when they decided to build a very large object system. They
mandated the use of the data model for identifying objects, resulting in
a class library of more than 5000 classes. This became the foundation
of their system, causing enormous implementation problems. A final,
midrange, example was a database application for billing and invoicing
wherein management demanded 1:1 replication of the existing system.
This was accomplished, but it took more than a year with an offshore
development team of 10 or 15 developers. My colleague and I dupli-
cated the capabilities of the system, using object thinking, in a weekend.
Management, however, was not impressed.

What criterion should be used instead of data or functions? Behavior!

Coad and Yourdon5 claimed that people have natural modes of thought.
Citing the Encyclopedia Britannica, they talk about three pervasive human meth-
ods of organization that guide their understanding of the phenomenological
world: differentiation, classification, and composition. Taking advantage of those
“natural” ways of thinking should, according to them, lead to better decomposition.

Behind the Quotes

Ed Yourdon and Peter Coad
Edward Yourdon is almost ubiquitous in the world of software development—
publishing, consulting, and lecturing for decades on topics ranging from
structure development to various kinds of crises (for example, the demise of
the American programmer and Y2K).

(continued)

5. Yourdon, Edward, and Peter Coad. Object Oriented Analysis. Yourdon Press. Englewood Cliffs, NJ:
Prentice Hall, 1990.
76 Object Thinking

Behind the Quotes (continued)

The foundation for his reputation arose from his popularization of

structured approaches to analysis and design. His textbook on structured
analysis and design was a standard text through several editions. In 1991,
he published two books, a new edition of Structured Analysis and Design
and a small book, coauthored with Peter Coad, called Object Oriented
Analysis.
In the object book, Yourdon made a surprising admission: the multi-
ple model approach—data in the form of an entity relation diagram, pro-
cess flow in the form of a data flow diagram, and implementation in the
form of a program structure chart—advocated in his structured develop-
ment writings (including the one simultaneously published) never, in his
entire professional career, worked! In practice, it was impossible to recon-
cile the conceptual differences incorporated into each type of model.
Objects, he believed, would provide the means for integrating the
multiple models of structured development into one. Unfortunately, he
chose data as the “knife” to be used for object decomposition. Other ideas
advanced in that book proved to be more useful for understanding objects
and object thinking—especially the discussion of natural modes of thought.
Peter Coad parted ways with Yourdon after the publication of this
book and developed a method and an approach to object modeling and
development that was far more behavioral in its orientation. He has several
books on object development that are worthy of a place in every object pro-
fessional’s library.

Classification is the process of finding similarities in a number of things

and creating a label to represent the group. This provides a communication and
thinking shortcut, avoiding the need to constantly enumerate the individual
things and simply speak or think of the group. Six different tubular, yellow, and
edible things become “bananas,” while five globular, red, edible things become
“apples.” The process of classification can continue as we note that both apples
and bananas have a degree of commonality that allows us to lump them into an
aggregate called “fruit.” In continuing the process of classification, we create a
taxonomy that can eventually encompass nearly everything—the Linnaean tax-
onomy of living things (and its more sophisticated DNA-based successors) being
one commonly known example.
Composition is simply the recognition that some complicated things consist
of simpler things. Ideally, both the complicated things and the simple things they
are composed of have been identified and classified. Grady Booch suggested
 Chapter 3 From Philosophy to Culture 77

that all systems have a canonical form. His book Object Oriented Design includes
a diagram captioned, “Canonical Form of Complex Systems,” which captures both
classification and composition hierarchies and the relationship that should exist
between the two.
Classification requires differentiation, some grounds for deciding that one
thing is different from another. The differentiation grounds should reflect natural
ways of thought, as do classification and composition. So how do we differentiate
things in the natural world?
Consider a tabby and a tiger. What differentiates a tiger from a tabby? Why
do we have separate names for them? Because one is likely to do us harm if
given the chance, and the other provides companionship (albeit somewhat
fickle). Each has at least one expected behavior that differentiates it from the
other. It is this behavior that causes us to make the distinction.
Some (people who still believe in data, for example) would argue that tab-
bies and tigers are differentiated because they have different attributes. But this is
not really the case. Both have eye color, number of feet, tail length, body mark-
ings, and so on. The values of those attributes are quite different—especially
length of claw and body weight—but the attribute set remains relatively constant.
Behavior is the key to finding the natural joints in the real world. This
means, fortunately, that most of our work has already been done for us. Soft-
ware developers simply must listen to domain experts. If the domain expert has
at hand a name (noun) for something, there is a good chance that that some-
thing is a viable, naturally carved, object.

Note Listening to the domain expert and jotting down nouns cap-
tures the essence of finding a natural decomposition. A full and shared
domain understanding requires the negotiation of a domain language,
a term used by Eric Evans in his forthcoming book Domain-Driven
Design: Tackling Complexity in the Heart of Software.

Using behavior (instead of data or function) as our decomposition crite-

rion mandates the deferral of much of what we know about writing software
and almost everything we learned to become experts in traditional (structured)
analysis and design. That knowledge will be useful eventually, but at the outset
it is at best a distraction from what we need to accomplish. We must relearn
how to look at a domain of interest from the perspective of a denizen (user) of
that domain. We need to discover what objects she sees, how she perceives
them, what she expects of them, and how she expects to interact with them.
Only when we are confident that our understanding of the domain and of its
78 Object Thinking

decomposition into objects mirrors that of the user and the natural structure of
that domain should we begin to worry about how we are going to employ that
understanding to create software artifacts. (Our understanding may come one
story at a time, à la XP.)

Note The focus of decomposition is understanding the domain as it is.

Developers and domain experts should always be aware that “what is”
is not necessarily “what is best.” Just because an object exists in the
domain in a particular form and has specific expectations associated
with it doesn’t mean that the object should and must continue to exist in
that form. Domains are subject to redesign, as are the objects and the
relationships and communications among objects in that domain. As
developers and domain experts work together, it’s quite possible that
they will define new objects and redesign existing objects. This is not
only acceptable but highly desirable—as long as the basis for redesign
activities remains the domain, not implementation environments.

Three: Objects must be composable.

As Plato noted, putting things together again is just as important as taking them
apart. In fact, it is the measure of how well you took them apart. Any child with
a screwdriver and a hammer can take things apart. Unless another child can
look at the pieces and determine how to put them together again (or even more
important, see how to take a piece from one pile and use it to replace a piece
missing from another pile), the first child’s decomposition was flawed.
Composability incorporates the notions of both reusability and flexibility
and therefore implies that a number of requirements must be met:

■ The purpose and capabilities of the object are clearly stated (from
the perspective of the domain and potential users of the object), and
my decision as to whether the object will suit my purposes should be
based entirely on that statement.
■ Language common to the domain (accounting, inventory, machine
control, and so on) will be used to describe the object’s capabilities.
■ The capabilities of an object do not vary as a function of the context in
which it is used. Objects are defined at the level of the domain. This
does not eliminate the possible need for objects that are specialized to
a given context; it merely restricts redefinition of that same object when
 Chapter 3 From Philosophy to Culture 79

it is moved to a different context. Objects that are useful in only one

context will necessarily be created but should be labeled appropriately.
■ When taxonomies of objects are created, it is assumed that objects
lower in the taxonomy are specialized extensions of those above them.
Specialization by extension means that objects lower in the taxonomy
can be substituted for those above them in the same line of descent.
Specialization by constraint (overrides) might sometimes be required
but almost inevitably results in a “bad” object because it is now impos-
sible to tell whether that object is useful without looking beyond what
it says it can do to an investigation of how it does what it says it can do.6

Although relatively simple to state, these requirements are difficult to satisfy.

The general principle guiding the creation of composable objects is to discover and
generalize the expected behavior of an object before giving any consideration to
what lies behind that behavior. This is a concept that has been a truism in computer
science almost from its inception. The most pragmatic consequence of this princi-
ple is the need to defer detailed design (coding) until we have a sure and complete
grasp of the identification and expected behaviors of our objects (the objects rele-
vant to the story we are currently working on) in the domain where they live.

Forward Thinking
A Problem of Reuse
In “Forward Thinking: Metaphor and Initial Stories,” which appeared in
Chapter 2, it was noted that two stories dealt with dispensing (change and
product) and might involve the same objects. Further discussion of dis-
pensing revealed three variations of a story involving some kind of
dispense action: dispense a measured volume of liquid, dispense a product,
and dispense change due the customer.
It would be nice if we had a single class, Dispenser, that could be
used in all three stories. This would mean that Dispenser would have to be
a composable object, able to be reused in different contexts without mod-
ification of its essential nature.

(continued)

6. The exception occurs when a method is declared high in the hierarchy with the explicit intent that all
subclasses provide their own unique implementation of that method and when the details of how are
idiosyncratic but irrelevant from the perspective of a user of that object.
80 Object Thinking

Forward Thinking (continued)

Because of the work of the team of developers on three different sto-

ries involving dispensing, three versions of the Dispenser object have been
created. The pairs of developers involved meet to look at each other’s
code and see whether they can refactor and redesign the Dispenser object
to make it more composable—more reusable.
In one case (product dispensing), the code for the dispense method
looked like the following pseudocode:
IF dispenserType = "Gate"
Gate open.
Else
Set timer = 10.
Open switch.
End-if
When timer =< 0 close switch.

The code in question reveals an awareness of two types of vending

evident in the machines in the hall: opening a gate to drop a can of soda
and pushing a product out of a coil.
The dispense method for the change dispenser looked like the
following:
While amountToBePaid >= SmallestDenominationAvailable
AND
AmountToBePaid > LargestDenominationAvailable
LargestDenominationDispenser ejectCoin
AmountToBePaid = (AmountToBePaid – largestDenomination).

Yes, I know you would never write code this ugly and that the sec-
ond example will not really work, but code is not the issue here; refactor-
ing is. After some discussion, the teams decided that the dispenser object
was really just a façade for some mechanism that did the actual work of
dispensing: a valve that opened for a period of time, a motor that ran for
a period of time, or a push bar that kicked an item out of the dispenser
storage area. It was also decided that the quantity to be dispensed should
be supplied to the dispenser rather than calculated by the dispenser.
These decisions simplified the method dramatically. In all cases, the
pseudocode would look something like this:
For 1 to quantityToBeDispensed
DispensingMechanism dispense.
End-loop.
 Chapter 3 From Philosophy to Culture 81

Forward Thinking (continued)

The only other behaviors of Dispenser—to disable itself when empty

or when not functioning and to identify itself—were already simple and
common in all contexts. Dispenser was now defined in such a way as to
be truly composable.
Was this accomplished only at the expense of moving some essential
complexity to another object? No. Two other objects are probably involved
in every dispensing operation: a collection object that contains the actual
dispensers and relays dispense requests to the appropriate dispenser
within the collection—a trivial behavior already built into well-designed
collection objects—and a dispensingRule object, which is an instance of
(not a subclass of) a SelfEvaluatingRule object. (See “Forward Thinking:
Communication and Rules,” for more discussion of rules in the UVM.)

Four: Distributed cooperation and communication must replace

hierarchical centralized control as an organizational paradigm.
Consider one of the more widely used models in traditional software development,
the program structure chart (Figure 3-1), popularized by Meillor Page-Jones.7 At
the top of the chart is the puppet master module, attended to by a court of special-
purpose input, transform, and output modules. The puppet master incorporates all
the knowledge about the task at hand, the capabilities of each subordinate module,
and when and how to invoke their limited capabilities. The same thinking charac-
terizes structured source code, wherein a main-line routine (frequently a Case
statement) consolidates overall control. Each paragraph of a collection of special-
purpose subroutine paragraphs is individually invoked and given limited authority
to perform before control reverts to the main line.
Unlike puppet modules, objects are autonomous. They are protected from
undue interference and must be communicated with, politely, before they will
perform their work. It is necessary to find a different means to coordinate the
work of objects, one based on intelligent cooperation among them.
It is sometimes difficult to conceive how coordination among autonomous
objects can be achieved without a master controller or coordinator. One simple
example is the common traffic signal. Traffic signals coordinate the movement
of vehicles and people but have no awareness of what those other objects are
about or even if any of them actually exist. A traffic signal knows about its own
state and about the passage of time and how to alter its state as a function
of elapsed time. In this model, the necessary “control” has been factored and

7. Page-Jones, Meillor. The Practical Guide to Structured Systems Design. Yourdon Press Computing
Series. Englewood Cliffs, NJ: Prentice-Hall, Inc.1988.
82 Object Thinking

distributed. The traffic signal controls itself and notifies (by broadcasting as a
different color) others of the fact that it has changed state. Other objects, vehi-
cles, notice this event and take whatever action they deem appropriate accord-
ing to their own needs and self-knowledge.

Master Control
Module

Afferent Transform Efferent

Module Module Module

Afferent Transform Transform Efferent

Module Module Module Module

Control Flow Data Flow

Figure 3-1 Program structure chart.
F03MQ01

Note But what about intersections with turn arrows that appear only
when needed? Who is in control then? No one. Sensors are waiting to
detect the “I am here” event from vehicles. The traffic signal is waiting
for the sensor to detect that event and send it a message: “Please add
turn arrow state.” It adds the state to its collection of states and pro-
ceeds as before. The sensor sent the message to the traffic signal
only because the traffic signal had previously asked it to—registered
to be notified of the “vehicle present” event. Traffic management is a
purely emergent phenomenon arising from the independent and
autonomous actions of a collectivity of simple objects—no controller
needed. If you have a large collection of traffic signals and you want
them to act in a coordinated fashion, will you need to introduce control-
lers? No. You might need to create additional objects capable of
obtaining information that individual traffic signals can use to modify
themselves (analogous to the sensor used to detect vehicles in a turn
lane). You might want to use collection objects so that you can conve-
niently communicate with a group of signals. You might need to make
a signal aware of its neighbors, expanding the individual capabilities of
a traffic signal object. You will never need to introduce a “controller.”
 Chapter 3 From Philosophy to Culture 83

Eliminating centralized control is one of the hardest lessons to be learned

by object developers.

Object Principles—Software Principles

Stating and explaining object presuppositions is important. It is also important
to show the relationship between those principles and generally accepted prin-
ciples of software design criteria. Exploring that relationship will further explain
and illustrate the object principles and show how they recast thinking about
design without rejecting traditional design goals.
Witt, Baker, and Merritt have written an excellent encapsulation of the
fundamental ideas about software design and architecture.8 Chapter 2 of their
book identifies a set of generally accepted axioms and principles that define
software quality:

■ Axiom of separation of concerns Solve complex problems by

solving a series of intermediate, simpler problems.
■ Axiom of comprehension Accommodate human cognitive limitations.
■ Axiom of translation Correctness is unaffected by movement
between equivalent contexts.
■ Axiom of transformation Correctness is unaffected by replace-
ment with equivalent components.
■ Principle of modular design Elaborates the axiom of separation
of concerns.
■ Principle of portable designs Elaborates the axiom of translation.
■ Principle of malleable designs Provides the means for composi-
tional flexibility.
■ Principle of intellectual control Appropriate use of abstractions.
■ Principle of conceptual integrity Suggests a limited set of concep-
tual forms.

Few would argue with these axioms and principles, although they would
certainly argue about the appropriate means for realizing them. Object thinkers
strive to achieve the goals implied by these axioms and principles as much as

8. Witt, Bernard I., F. Terry Baker, and Everett W. Merrit. Software Architecture and Design: Principles,
Models, and Methods. Van Nostrand Reinhold, 1994.
84 Object Thinking

any other software developer and believe that objects provide the conceptual
vehicle most likely to succeed.
For example, the separation-of-concerns axiom and the principle of mod-
ularity mandate the decomposition of large problems into smaller ones, each of
which can be solved by a specialist. An object is a paradigmatic specialist. Large
problems (requiring a number of objects, working in concert to resolve
the problem) are decomposed into smaller problems that a smaller community
of objects can solve, and those into problems that an individual object can deal
with. At the same time, each object addresses the principles of intellectual con-
trol (individual objects are simple and easy to understand) and the principle of
conceptual integrity (there should be a small number of classes). Properly con-
ceived, an object is a natural unit of composition as well. An object should
reflect natural, preexisting decomposition (“along natural joints”) of a large-
scale domain into units already familiar to experts in that domain. Conceived in
this fashion, an object clearly satisfies the principle of intellectual control.
Objects will also satisfy the principle of conceptual integrity because there will
be a limited number of classes of objects from which everything in the domain
(the world) will be constructed. In Chapter 4, “Metaphor: Bridge to the Unfa-
miliar,” an argument will be presented suggesting that the total number of
objects required to build anything is around 1000.
Objects are designed so that their internal structure and implementation
means are hidden—encapsulated—in order to satisfy the axiom of transforma-
tion and the principle of portable designs.
The principle of malleable designs has been the hardest one for software to
realize: only a small portion of existing software is flexible and adaptable enough
to satisfy this principle. In the context of object thinking, malleability is a key moti-
vating factor. Object thinkers value designs that yield flexibility, composability, and
accurate reflection of the domain, not machine efficiency; not even reusability,
although reusability is little more than cross-context malleability.
In fact, it might be said that for object thinkers, all the other axioms and
principles provide the means for achieving malleability and that malleability is
the means whereby the highest-quality software, reflective of real needs in the
problem domain, can be developed and adapted as rapidly as required by
changes in the domain. Agile developers and lean developers9 value malleabil-
ity as highly as object thinkers. XP software systems emerge from software that
satisfies the demands of a single story, an impossibility unless it is easy to refac-
tor, adapt, and evolve each piece of software (each object); impossible unless
each bit of software is malleable.

9. Poppendieck, Mary, and Tom Poppendieck. Lean Software Development: An Agile Toolkit for Software
Development Managers. Addison-Wesley. 2003.
 Chapter 3 From Philosophy to Culture 85

Fred Brooks wrote one of the most famous papers in software develop-
ment, “No Silver Bullet: Essence and Accidents of Software Engineering.”10 In
that paper, he identified a number of things that made software development
difficult and separated them into two categories, accidental and essential.
Accidental difficulties arise from inadequacies in our tools and methods and
are solvable by improvements in those areas. Essential difficulties are intrinsic to the
nature of software and are not amenable to any easy solution. The title of Brooks’s
paper refers to the “silver bullet” required to slay a werewolf—making the meta-
phorical assertion that software is like a werewolf, difficult to deal with. Software,
unlike a werewolf, cannot be “killed” (solved) by the equivalent of a silver bullet.
Brooks suggests four essential difficulties:
■ Complexity Software is more complex, consisting of more unlike
parts connected in myriads of ways, than any other system designed
or engineered by human beings.
■ Conformity Software must conform to the world rather than the
other way around.
■ Changeability A corollary of conformity: when the world changes,
the software must change as well, and the world changes frequently.
■ Invisibility We have no visualization of software, especially exe-
cuting programs, that we can use as a guide for our thinking.
He also investigates potential silver bullets (high-level languages, time
sharing, AI, and so on) and finds all of them wanting. Object-oriented pro-
gramming is considered a silver bullet and dismissed as addressing accidental
problems only.
Although I would agree with Brooks in saying that object technology—
languages, methods, class hierarchies, and so on—addresses only accidental
problems, object thinking does address essential difficulties, and it does so with
some promise. Objects can conform to the world because their design is pred-
icated on that world. Objects are malleable, resolving the changeability issue.
Objects provide a way to deal with the complexity issue and even allow for the
emergence of solutions to complex problems not amenable to formal analysis.
The metaphors presented in Chapter 4 provide the tools for visualization to
guide our thinking.
Object thinking suggests we deal with software complexity in a manner
analogous to the ways humans already deal with real-world complexity—using
behavior-based classification and modularization. Object thinking is focused
on the best means for dealing with conformity and changeability issues—the

10. IEEE Computer, April 1987.

86 Object Thinking

malleability principle—as a kind of prime directive. And invisibility is

addressed, not with an abstract geometry as suggested by Brooks, but via simu-
lation (working software using an XP perspective)—direct, albeit metaphorical,
simulation of the real world. If we can understand the complex interactions of
objects in the real world (and we do so every day), we should be able to visualize
our software as an analogous interaction of objects.

Forward Thinking
Communication and Rules
Because the UVM might be dispensing food items and because we want
the customer experience to be always positive, we want to ensure that no
spoiled products are vended. This leads to a story—Expire: no product is
sold after its expiration date has been reached.
The development team discusses (and codes) various ways this
might be accomplished. Through a combination of refactoring efforts and
arguments, it is decided that the expiration problem will best be solved by
a group of objects communicating with one another, with those commu-
nications being triggered by events.
Whenever a product is placed in the vending machine, it asks itself
for its expiration date. It then asks the SystemClockCalendar to add an
eventRegistration (consisting of the productID and the “die” message) for
the event generated whenever a new day is recognized by the System-
ClockCalendar. (Programmers, even extreme programmers, often have a
rather grim sense of humor; hence the “die” message to effect product
expiration.) At the same time, the Dispenser object asks the new product
to accept a registration for the “I’m dead” event that the product will gen-
erate when it receives the “die” message from its own event registration that
was placed with the SystemClockCalendar. The dispenser’s eventRegistration
with the product will cause the message “disableYourself” to be sent to the
dispenser, who will, indeed, “disable” itself (with the accompanying event
that other objects—such as the menu or the dispenser collection—might
register for).
Breaking up a potentially complex decision-making and cascading-
effects problem into pieces that can be distributed among many objects
while at the same time relying on simple, reusable components such as an
eventRegistration and a Dispatcher greatly reduces the complexity that
worries Brooks. It also accommodates the conformity and changeability
requirements imposed on software: event registrations can be added or
 Chapter 3 From Philosophy to Culture 87

Forward Thinking (continued)

deleted as needed, redirected to other objects, or transformed so that the

registering object receives different messages at different times without the
need to rewrite and recompile source code.
The development team found another opportunity for simplification as
they worked with various types of rules that governed actions in different
parts of the UVM. One kind of rule was, “Don’t vend a product unless suf-
ficient funds have been accumulated.” Another rule was, “Refund money
using the largest coins available, moving to lower-denomination coins only
when the larger denomination is greater than the sum yet to be refunded.”
Using a combination of refactoring and appropriate abstraction, the
development team defined and designed a rule object. (XP philosophy
warns against premature abstraction: abstraction that is not derived from
refactoring, meaning not grounded in efforts to achieve simplification.
Hence the adjective appropriate in the preceding sentence.) A rule is an
ordered collection of constants, variables, and operations. A variable con-
sists of an object and a message to be sent to that object. When a variable
sends the message to the target, the resultant value replaces the unknown
value of the variable. When asked to evaluate to a result, a rule iterates
across its elements, asking each variable to instantiate itself to a real value,
and then applies the operators to the instantiated variables and constants.
All four of Brooks’s concerns about software’s essential difficulties
are addressed: simplification of complexity, ease of conformity and adapt-
ability, and visualization. A rule is an easy thing to visualize—we see
examples of them in everyday life frequently—and the process of instan-
tiation and resolution is very straightforward—we can see it operating in
our mind’s eye with no difficulty. Reliance on simulation constantly pro-
vides other visualizations of the software we are creating.

Cooperating Cultures
Arguing for the existence of an object paradigm or object culture is not and
should not be taken as an absolute rejection of traditional computer science
and software engineering11. It would be foolhardy to suggest that nothing of
value has resulted from the last fifty years of theory and practice.

11. The ideas in this section were first published in a short editorial by the author in Communications
of the ACM, 1997.
88 Object Thinking

Claiming that there are clear criteria for determining whether software is
object oriented is not the same as saying all software should be object oriented.
Expecting a device driver implemented with 100 lines of assembly language
to reflect full object thinking is probably pointless. It’s possible that Internet
search engines must be implemented using “database thinking” rather than object
thinking—at least for the immediate present. Different problems do require dif-
ferent solutions. The vast majority of problems that professional developers are
paid to deliver, however, almost certainly require object solutions.
Traditional approaches to software—and the formalist philosophy behind
them—are quite possibly the best approach if you are working close to the
machine—that is, you are working with device drivers or embedded software.
Specific modules in business applications are appropriately designed with more
formalism than most. One example is the module that calculates the balance of
my bank account. A neural network, on the other hand, might be more herme-
neutic and objectlike, in part because precision and accuracy are not expected
of that kind of system.
Traditional methods, however, do not seem to scale. Nor do they seem
appropriate for many of the kinds of systems being developed 50 years after the
first computer application programs were delivered. Consider the simple graph
in Figure 3-2.

Natural -
Sociocultural
World
Object
Paradigm

Comprehension
Implementation

Computer
Science
Paradigm
Deterministic
World

Figure 3-2 Applicability graph.

F03MQ02
 Chapter 3 From Philosophy to Culture 89

The horizontal axis represents the spectrum of activities involved in sys-

tems modeling and application development. It ranges from analysis (with the
accompanying tasks of comprehending the real world, making useful abstrac-
tions, and decomposition) to implementation (compiling, testing, and executing).
The vertical axis opposes the deterministic world (the domain of hardware,
discrete modules, algorithms, and small-scale formal systems) to the natural
world (businesses and organizations, societies, composite systems, and cultures).
A diagonal bisects the graph to demarcate two realms. To the lower right
is the realm where mainstream computer science and formalist ideas have dem-
onstrated success. Emphasis in this realm is on defining and building hardware
and using a finite set of representations (binary and operation codes) and
manipulation rules (the grammar of a compiler) to implement software designs.
This is the realm of formalism, where systems can be complicated and even
large but not complex—that is, they do not exhibit nondeterministic behavior,
they are not self-organizing, and they do not have emergent properties.
At the upper left is an area that is largely terra incognita as far as computer
scientists are concerned. This is the realm of social, biological, and other com-
plex (as that term is coming to be understood) systems. Meaning in this realm
is not defined; it is negotiated. Rules are not fixed but are contextual and
ephemeral. Constant flux replaces long-term consistency. This is the arena in
which hermeneutic and object ideas offer an expansion of our ability to model
and build systems capable of interacting with the natural systems in which they
will have to exist.
This is the realm where objects and the object paradigm should dominate.
Objects provide a foundation for constructing a common vocabulary and a
process of negotiated understanding of the complex world. Behavioral objects
offer a decomposition technique that will yield adaptable constructs for building
highly distributed, “intelligent,” and flexible computer artifacts and computer-
based systems.
Object thinking provides a foundation for attaining an integration of arti-
ficial systems (computer and software) with the natural systems (social and
cultural contexts). Object thinking requires an awareness of the domain and the
fitness of our artifacts as they operate in that domain in ways that traditional
thinking (“Make the artifact meet specification”) cannot. Object thinking cou-
pled with the values of XP, especially communication, create a basis for true
collaboration among users, managers, and developers.
All of this, once we learn object thinking.
Metaphor: Bridge to
the Unfamiliar
Along the philosophical fringes of science we may find reasons to question
basic conceptual structures and to grope for ways to refashion them. Old
idioms are bound to fail us here, and only metaphor can begin to limn the
new order.
—Willard Van Orman Quine

Explanations without metaphor would be difficult if not impossible, for in

order to describe the unknown, we must resort to concepts that we know and
understand, and that is the essence of metaphor.
—Earl MacCormac

Object philosophy generates a different view of the world, one that is strange to
most and especially to developers trained in conventional methods and ideas.
This raises the question of how best to assist developers to understand the new
world of objects. In other disciplines, metaphor is frequently used to help those
new to an area of study comprehend its fundamental concepts.
Consider the Bohr model of an atom, as one example. Physicists and
chemists need to explain atomic structure to lay people and to new students.
One common way to do so is to employ Bohr’s metaphoric model that says an

91
92 Object Thinking

atom is like a tiny solar system—a nucleus (sun) surrounded by orbiting electrons
(planets). This metaphor is technically inaccurate, of course, but it remains a useful
tool for introducing atomic concepts.
Unsurprisingly, metaphors have also been used to convey object concepts.
One of the earliest, coined by Brad Cox, is the software integrated circuit (IC).
This metaphor juxtaposes a desirable trait for software objects with hardware
components, that is, the ability to use them as standardized and interchangeable
parts to “mass-produce” larger constructs. It’s possible, for example, to shop at a
number of electronics stores, buy standard components from a variety of manu-
facturers, and use those components to assemble a working personal computer.
Construction of software in a similar manner is one goal of object orientation;
hence the applicability of the metaphor.

Behind the Quotes

Brad Cox
Brad Cox was one of the earliest advocates of object-oriented program-
ming and was the developer of the Objective-C programming language,
which was the core of the Next Computer operating system and develop-
ment environment. (Next was the first company started by Steve Jobs
when he left Apple Computer.) Objective-C extended the standard C pro-
gramming language with Smalltalk characteristics—an attempt to realize
the expressiveness and simplicity of Smalltalk while retaining the efficiency
and power of C.
Dr. Cox coined an early metaphor for object components—software
ICs—suggesting that software components should be as modular and as
composable as the ICs used to build hardware. Given an appropriate set of
ICs, software could be mass-produced—cheaply and with quality—just as
guns were after Colt and Remington invented standardized parts for guns.
One of Dr. Cox’s most intriguing ideas was the concept of superdistri-
bution. The core idea of superdistribution was to make individual software
modules available via the Web and charge for their use, component by
component—a concept that is evident in Microsoft’s vision of Web services.
Dr. Cox recently resigned from George Mason University to concentrate
on a company that will make superdistribution a reality. He has many other
interests, and his Middle Of Nowhere (http://www.virtualschool.edu/mon) is
well worth a visit.
 Chapter 4 Metaphor: Bridge to the Unfamiliar 93

The software IC metaphor is less helpful, however, when our concern is

object discovery and specification because it tells us only of a desirable trait for
the finished product. Alan Kay, Adele Goldberg, Ken Rubin, and Philippe Kahn,
among many others, employ the metaphor of a person when engaged in the
process of discovery and analysis. An object is like a person. This metaphor is
sufficiently important to object thinking that it warrants its own special term—
anthropomorphization. And of course, the more syllables a word has, the more
important it must be. We will return to this metaphor later.
Metaphors are not just a tool to explore the unfamiliar. Metaphor is essen-
tial to everyday thinking as well. The full importance of metaphor in shaping
our thoughts has received a lot of attention in recent years, notably in the work
of George Lakoff.

Behind the Quotes

George Lakoff
Professor Lakoff teaches and researches linguistics, cognitive science, and
cognitive philosophy at the University of California, Berkeley. His books
most likely to be of interest to readers of this book include Metaphors We
Live By; Women, Fire, and Dangerous Things; and Philosophy in the Flesh.
Lakoff’s work is frequently cited by those criticizing traditional approaches
to software development—especially the kind of set-based category the-
ory underlying traditional approaches to data and data modeling—as well
as by those advocating a more human-centric approach to computing and
the software development process.
Dr. Lakoff’s research reveals the central role of metaphor in all human
cognition and provides a foundation for Kent Beck’s ideas about metaphor
in extreme programming—everything from the value of a system metaphor
(in lieu of architecture) to the need for metaphorical awareness when creat-
ing object and method naming conventions.

Metaphor shapes our thinking in many different ways.

■ It helps in discovery. “If our system is like an X, this component

is probably like a Y.”
■ It helps us make design decisions. “When a person is asked for
ID, she usually hands over some sort of document (such as a driver’s
license), so our Person probably should store identifying information
in a similar kind of object instead of as a sequence of instance vari-
ables containing single values."
94 Object Thinking

■ It provides handy ways to remember principles of object

thinking. “Objects are naturally lazy, and this is starting to look
hard. We had better refactor our design and split this work up among
several objects.”
■ It helps us avoid old ways of thinking by avoiding the metaphors
that are associated with those kinds of thinking. “Instead of ‘next
the machine needs to do this,’ we use, ‘Just ask object X to do that.’”

Metaphor plays a critical role in XP as well as in object thinking. Kent

Beck used his keynote address at OOPSLA 2002 to explore all the ways that
metaphor affects all aspects of XP. One of the 12 practices in XP is the system
metaphor, which is deemed powerful enough to eliminate the need for detailed
up-front architectural design to guide development. In Kent’s book on test-
driven development,1 he talks about how different metaphors led to several dif-
ferent designs and implementations of multicurrency money.

Forward Thinking
A Vignette
Roger and Suroor are working on the “accumulate money” story, the user
story about how the vending machine accepts coins and currency and
reports back a total. They have test cases based on random selection from
among all the possible denominations of coins and paper money from all
three target currencies (euros, yen, and dollars) and are starting to code
the Accumulator class.
“I guess we will need a method to determine what kind of currency we
have and what its value is,” suggests Roger, “and one to convert the curren-
cies into some kind of common value—all dollars, or yen, or euros. A method
to report the total amount in the accumulator and one to actually add each
instance of inserted money to the current accumulation,” he continued.
“Don’t forget some kind of reset-to-zero method,” added Suroor.
“Two of these methods are really easy.” Suroor is talking while typing
code for two methods: reset −total = 0; and total, which just returns the
amount in the instance variable named total.
“Determining the kind and value of the currency looks like it might
be complicated,” says Roger, reaching for the keyboard, “some kind of
case statement probably.”

1. Beck, Kent. Test Driven Development by Example. Boston Addison-Wesley, 2003. ISBN 0-321-14653-0.
 Chapter 4 Metaphor: Bridge to the Unfamiliar 95

Forward Thinking (continued)

“No, no,” interjected Suroor. “Remember, object code shouldn’t need

case statements—we need to rethink which object is responsible for iden-
tifying currency type and value. When I was pairing yesterday on the
credit card story, we created a Money class that identified its value and its
currency. We should use that.”
“You’re right; some way or another, every bit of money put in the
machine identifies itself with a currency type and a value,” Roger agreed,
nodding. “After all, that is what happens now—somehow money uses
inertia and shape to tell the machine what it is. Maybe someday there will
be a chip in coins and currency that does the job easier—but whatever, it
is the money’s responsibility.”
“That makes the add method real easy,” said Suroor as she typed the
new method: amount in total, add amount inserted.
“But wait, what about currency conversion?” Suroor was talking out
loud to herself.
“Money should do that too,” suggested Roger. “All money objects should
have methods to convert themselves into other currencies—something like
asEuro or asDollar.”
“The team creating the Money object is doing that—using a currency
conversion lookup table for the actual exchange rate—a collaborator for
the money object.”
“How do we know what currency to convert to?”
“Ask the VendingMachine object—it will be a global object that can
be used to store bits of information that all the other objects in the vend-
ing machine application might need to access. It can have a variable
named localCurrency or defaultCurrency. Make a note.”
“What does the code look like now?”
“Well, we only have three methods—reportTotal, resetTotal, and
addToAccumulation. The first two are just one line of code each. The last
one is more complicated, but not much. The method name is addToTotal,
and it has an argument, moneyInserted. The actual method has three levels—
see how it is nested:
total addToTotal (moneyInserted as (vendingMachine.localCurrency)).

“Looks like we kept our object appropriately lazy—making the money

object do all the work of identifying itself, conversion, and addition.”
“Yeah, but those methods are easy for money to do for itself—we
actually simplified the system by doing this.”
96 Object Thinking

Metaphors are very powerful. Object thinking is absolutely dependent on

selecting and employing the “right” metaphors. Each metaphor—whether general
or specific to design details—must be consistent with the philosophy behind
objects as discussed in previous chapters. In this chapter, we will introduce several
key metaphors.

The Lego Brick Metaphor

Our first metaphor helps us think of important object characteristics, such as
composability, simple interfaces, and comprehensibility (limited number of
forms). It also illuminates important aspects of the object-oriented software
development process, most notably the fact that two distinct but related pro-
cesses are required.
Let’s express the metaphor as a dictum: software should be assembled from
a finite set of composable units the way that dinosaurs and castles and space-
ships are constructed from a common set of Lego bricks. It is no accident that the
first special issue of the Communications of the ACM devoted to objects had a
cover photo of stacked red and yellow Legos.
On its face, this seems to be a restatement of Brad Cox’s software IC met-
aphor. But there are depths and nuances to this metaphor that are missing from
the simpler software IC idea. Like Cox’s metaphor, this one tells us that we
should be able to construct an arbitrary number of software artifacts from a
finite set of standard parts. This is a characteristic of many other aspects of
nature as well as of mass-produced products. For example, the world around us
is constructed from a very finite set of parts—atoms of the periodic table or,
below them, quarks—especially given the variety and complexity of the things
manifest in the world. Houses are constructed from two-by-fours, nails, 4-by-8-
foot sheets of plywood, and so forth. Even societies are built from “butchers,
bakers, and candlestick makers.”
By focusing on the composable nature of Lego bricks, the metaphor reminds
us of how object thinking views the importance of decomposition and composi-
tion. (Remember Plato’s views, quoted earlier in this book.) The metaphor also
reminds us of the importance of composability across contexts, a generalization of
the concept of reuse. Removing a Lego brick from a dinosaur and using it to shore
up the antenna on your space station is an example of composability (reuse)
across contexts.
More important, perhaps, the metaphor reminds us that simple and obvi-
ous interfaces are required if our objects are to be as composable and as useful
as Lego bricks. (The standard physical dimensions of a Lego brick are but
another aspect of its public interface.) Further exploration of the metaphor and
 Chapter 4 Metaphor: Bridge to the Unfamiliar 97

its implications suggests ways to discover, design, and build truly reusable and
composable objects. This alone would make the metaphor extremely valuable.
Given that developers have pursued the dream of reusable code libraries from
the very advent of computing—with very limited success—a way to actually
accomplish that goal would be invaluable.

Note Almost all software development has focused on a very limited

concept of reuse instead of the composability implicit in the axiom of
Witt, et al. and Plato’s philosophy of “taking apart and putting together
again.” Some would claim that the industry track record with reusable
libraries, class libraries, components, and widget libraries suggests
that reuse across contexts is impossible. At issue, however, is not the
possibility of composability but the means of achieving it. Reuse
efforts (as opposed to composability efforts, which have been almost
nonexistent) have been characterized by a focus on the solution
space—computer code, algorithms, and data structures. True com-
posability will require an understanding of the problem space—of the
natural world as advocated by object thinking. In Chapter 10, we will
explore the issue of composability further in a discussion on the future
of objects.

Exploration of the metaphor begins with considering what it implies about

the development process: that there is a necessary separation between the pro-
cess required to create objects (Lego bricks) and the process of assembling those
objects into useful products (software, in our case). The metaphor suggests that

■ Creators are (probably) adults working for the Lego Company.

■ Users are children (at least at heart), ages 4 to adult.
■ Creators have specific concerns and use specialized processes to
accomplish their goals.
■ Users care little about the components, as components. Their con-
cerns focus on what can be built with the components and the ability
of the artifact to satisfy their needs.
■ The component “engineers” need to be very concerned with the
internal structure of the bricks, what kind of plastics will yield the
correct degree of malleability, colorfastness, friction to keep them
together, spacing of the pips at the top of the brick, and so on. They
98 Object Thinking

also have to create components that transcend particular applications

because their goal is to build bricks equally useful for dinosaur and
spaceship construction. More important, they want bricks that can be
used successfully by unknown end users to build whatever it is that
they have imagined.
■ Users want to move easily from concept to construction without the
need to concern themselves with technical details. They want rapid
feedback, they want to be able to change their mind in mid-construction,
and they want the artifact constructed to operate in their world as it is.

Applied to software, the metaphor suggests separation of domain decom-

position and object definition from the tasks of assembling applications and
solving specific operational problems. To some degree, with some implemen-
tation languages, this separation has started to occur. Consider the class libraries
that come with a language such as Smalltalk or Visual Basic. Many of the classes
(the collection and magnitude classes, for instance) in such libraries reflect the
same kind of general and abstract thinking that leads to good “software Legos.”
Another example of this separation is the attempt to create visual programming
environments for application assembly that are at least quasi-independent of
the underlying implementation language. Such attempts are but a start toward
an object-mature world, where the two tasks are as clearly separated as they are
in the world of Legos.

Note I would like to bring to the reader’s attention a nuance of this

metaphor as something to think about without attempting to fully
develop the idea. The metaphor suggests differentiation between
users and creators that, in the case of the Lego brick, is very different
from the similar distinction made in this book. A Lego user is a child—
a kind of ultimate consumer. In the case of objects and software, we
are treating other programmers as users.
If we were to be completely consistent with the Lego metaphor,
we would have to argue in favor of delivering objects to end users,
those filling roles in business and organizational worlds, and not to
programmers. The “objects” would have to be directly usable without
the need to use programming environments and compilers. Each com-
ponent would need to be a small executable program, modifiable via
user messages, not modules of source code made available to pro-
grammers.
 Chapter 4 Metaphor: Bridge to the Unfamiliar 99

Another, critically important, aspect of the metaphor is the ability of users

to successfully employ the bricks based solely on their intuitively obvious exter-
nal characteristics. A child can look at a brick and instantly tell whether it is suit-
able for inclusion in the project at hand. It is not necessary to know anything
about the chemistry of plastics or whether this particular brick was made at
“Sun BuildingBlock Corporation” or “Microsoft BuildingBlock Corporation.”
There is no need to read a complex user manual that explains either the brick or
how to use it. There are, however, patterns: diagrams suggesting proven ways
that you can construct a family of similar artifacts—houses, for example.
Software objects cannot even approximate this degree of composability,
but the metaphor suggests that full realization of object potential requires satis-
faction of this characteristic.
The fact that two different groups of people, and two different processes,
are involved in brick creation and artifact assembly, one group being adults and
the other children, might lead one to believe that the metaphor de-skills the
task of application assembly—after all, it can be done by children (or end users,
perhaps). This would be a misleading conclusion.
The Legoland store in the Mall of America periodically sponsors two events.
(Located in Bloomington, Minnesota, the MOA is the largest shopping mall in the
United States.) In the first event, children are invited to use the unlimited set
of bricks at hand to construct various things. Prizes are given to the best con-
structions. The second invites professional architects and designers to use the
same bricks to create various structures, which are then sold as part of a charity
auction.
As should be expected, there was a large qualitative difference between
the constructions of the architects and those of the children. The building bricks
remained the same. The architects, however, were able to bring to bear other
skills—proportion, geometry, aesthetics, and so on—that the children did not
yet possess. The architects were domain experts (end users) who were able to
use the bricks to build solutions that fully exploited their domain knowledge.
They were able to use the Lego bricks to simulate the way they wielded girders
and bricks in the real-world domain where they worked.

Note Following up the immediately preceding note, perhaps the users

of objects would not have to be the ultimate end user but could be a new
kind of professional assembler or collage artist, with a set of skills not
available to the end user but quite different from those required by tradi-
tional programmers and software developers. The demand that objects
be run-time modifiable executable programs would not change.
100 Object Thinking

Two other items suggested by the metaphor are related. Objects should be
simple, and there should be relatively few of them. There are fewer than 10 basic
Lego brick types. (Kits contain additional parts, each of which is highly specialized,
such as tiny human figures and motors, but these cannot be considered true Legos.)
There are only 134 elements and only six quanta. The vast majority of houses in this
country are built with fewer than 10 standard sizes of dimension lumber. In all of
those cases as well, the base elements are simple and highly specialized.

How Many Objects?

One of the heuristics for object discovery is “Find the nouns.” Each noun
in a domain description is a potential object class. Estimating the minimal
class set could involve a simple count of nouns employed in a domain.
Expanding on this heuristic, how many classes would be required to
model the universe? Well, how many nouns are required?
The Oxford English Dictionary has about 550,000 words. An English
teacher once told me that about 40 to 45 percent of the words in a dictio-
nary will be nouns.
If we eliminate proper nouns and synonyms, that percentage will be
reduced to around 25 to 30 percent of the words describing potential
objects. Eliminating archaic nouns—bodkin and amanuensis, for example—
will reduce the percentage still further, to 15 or 20, perhaps. This translates
into about 110,000 classes. A pretty large number but clearly finite.
Instead of the OED, however, a better estimate might be obtained
using the vocabulary required to read the average daily newspaper. Most
things are quite adequately described in a newspaper.
Vocabulary required to read a typical newspaper: about 1400 words!
Using the 30 percent estimate (we don’t have archaic terms to elim-
inate, hopefully) suggests a need for only 420 classes. Allowing liberal
ability to add classes representing objects that do not make the paper, we
still come up with fewer than a thousand classes to model all typical
domains of human interest.
If your problem domain—or worse, your application—has thousands
of classes (and I have seen some), you probably have yet to master object
thinking.

This suggests that there are a similarly small number of objects from which
we can construct any type of software needed to model any domain or organi-
zation. (See the sidebar “How Many Objects?”)
 Chapter 4 Metaphor: Bridge to the Unfamiliar 101

It’s frequently convenient to build a large construct from small, but not the
smallest possible, components. It’s easier to build living things with hydrocar-
bon molecules than directly with individual atoms. It’s easier to build a roof
with a truss made of standard-dimension lumber and gang nails than a board
and a nail at a time. Intermediate constructs, such as trusses used to build
houses, are components. The number of components will be much larger than
the number of objects from which those components are constructed. More-
over, they will likely reflect stylistic differences reflective of the designers and
potential users of such components.
The last item suggested by the metaphor deals with process. Watching a
child work with the bricks reveals a process of discovery filled with a certain
amount of trial and error and supportive of rapid change as prototypes fail to
meet satisfaction criteria and so are taken apart and reassembled in another
attempt to reach the envisaged goal.
XP development closely resembles playing with Lego bricks in the sense
that it too allows discovery and emergent solutions, tolerates and leverages mis-
takes, encourages taking things apart and reassembling them into more elegant
solutions (refactoring), and relies heavily on feedback as to the extent to which
the current assembly meets user expectations.
Both XP and the metaphor suggest a need for a development environment
that supports this kind of development process model. Smalltalk and visual pro-
gramming environments, such as Visual Studio, provide examples of development
environments and tools that are superior to, in this regard, compile-link-test envi-
ronments such as C++ (even Visual C++ with incremental compilation). This notion
would seem to be borne out in experience. It typically takes about half as long to
develop an application with Smalltalk as with C++, given equivalent levels of skill
in the developers, even with current incremental compilers. Even advocates of lan-
guages such as C++ will tend to concede the speed-of-development issue and
focus instead on characteristics-of-product issues such as speed of execution.

The Object-as-Person Metaphor

People are objects, and objects should be conceptualized as if they were people.
Projecting human characteristics onto inanimate things is called anthropomor-
phization. Anthropomorphizing in a philosophy class might earn you a poor
grade, but it’s essential to a good understanding of objects.
Philippe Kahn starred in The World of Objects, a video he produced to
illustrate object concepts. He used musicians, a specialized type of human
being, to illustrate the anthropomorphization of objects. Mr. Kahn noted that he
is capable of playing numerous instruments and splicing the results together to
create a finished piece of music. Better results with less work, however, would
102 Object Thinking

be obtained by engaging a group of talented musicians—each of whom was a

specialist in some instrument or aspect of music. He could then communicate
his desires to them, allowing them to respond to those desires using their innate
skills, knowledge, and abilities. The musicians would be asked, at a relatively high
level, for some service (asking the percussionist for some “color,” for example).
They in turn would interpret (make sense of) those instructions in terms of their
own abilities, experience, and even awareness of the context in which the request
was made and then respond in an appropriate manner.

Behind the Quotes

Philippe Kahn
Philippe Kahn is the founder of Borland International, once one of the
largest software companies in the world and still a leader in the field of
integrated development environments (IDEs), which combine program-
ming language, editing, incremental compiling, debugging, and visual
tools to simplify and speed software development.
Borland was an aggressive early adopter of object ideas for its own
software development and enjoyed numerous early successes with object
technology. Mr. Kahn left Borland, and the world of the PC, to concentrate
on network applications as CEO of Starfish Software. His current work still
reflects a commitment to object ideas.

The object-as-person metaphor tells us that an object needs to be an agent

capable of providing a specified set of services. It has access to a body of
knowledge (some of it internalized) that it uses to respond to our service
requests as well as any necessary mechanisms (talents and skills) and resources
(instruments, a computer, or whatever else it may need). It also tells us that the
(only) appropriate way to determine whether an object will suit our needs is by
a careful review of its interface (résumé).
Applying this metaphor can feel a bit strange at first. What, for example,
are the talents or skills possessed by a book? What services can or does it provide?
What knowledge does it require to respond to requests for those services? What
resources does it need? These appear to be hard or nonsensical questions to
answer. But with some practice, a book is revealed as

■ A collection of page objects.

❑ Working with those page objects, it maintains order (sequence).
 Chapter 4 Metaphor: Bridge to the Unfamiliar 103

■ An object capable of identifying itself.

■ An object that can describe itself.
❑ The description is a separate object that can provide individual
elements of the description, such as date published, author or
authors, publisher, and so on.
■ An object that can provide the reader with access to a specific page
or a group of pages (chapters) upon request (also in collaboration
with the page or chapter objects).
■ An object that acts as a front for a community of objects (pages,
tables of contents, indexes, chapters, and the like), allowing users of
the community a convenient point of contact. Requests can be sent
to the book, knowing that the book will relay those requests to the
actual objects capable of responding, without interference.
■ At one point in its life (in the days when it was a mere unpublished
manuscript), an object that was also able to add pages to and delete
pages from its collection and still maintain the proper order.

If objects are similar to persons, they are limited in some of the same ways
that human persons are limited. For example, they need to know certain things to
complete an assigned task. When the book was young and still being composed
(playing the role of manuscript), it might have been asked to add a page to its col-
lection of pages. To do this, it would have needed to know the page to be added.
Like people, software objects are specialists. They are also lazy. A conse-
quence of both these facts is the distribution of work across a group of objects.
Take the job of adding a sentence to a page in a book. Granted, it might be
quite proper to ask the book, “Please replace the sentence on page 58 with the
following.” (The book object is kind of a spokesperson for all the objects that
make up the book.) It would be quite improper, however, to expect the book
itself to do the work assigned. If the book were to do that kind of work, it
would have to know everything relevant about each page and page type that it
might contain and how making a simple change might alter the appearance and
the abilities of the page object. Plus the page might be offended if the book
attempted to meddle with its internals.
The task is too hard (lazy object) and not the book’s job (specialist object),
so it delegates—merely passes to the page object named #58 the requested
change. It’s the page object’s responsibility to carry out the task. And it too
might delegate any part of it that is hard—to a string object perhaps.
Following these patterns, we let the book and the page do what each
does best and what is appropriate for each. If we need a service that requires a
104 Object Thinking

contribution from more than one object, either we assume responsibility for
asking the objects for their individual contributions and assembling those
results in a way that suits our purpose or we send the request to whichever
object is acting as spokesperson for the group and allow it to delegate tasks and
assemble responses in order to reply to our request.
The person metaphor guides our decomposition and our assignment of
responsibilities to software objects that always reflect the demands of the
domain, not their eventual implementation. Our software object should simu-
late the services provided by our real-world object, both of which we meta-
phorically regard as people. Even though it’s true that we must be more precise
in specifying our software object, it’s critically important that we continue to use
the real world and not the computer (implementation) world as the foundation
for our conceptualization of the software object.

Forward Thinking
Hey, Coach!
“Hey, coach,” Lori waved to Ron, who was playing the role of coach for
the teams today, “could you grab Hector [their on-site customer] and
mediate a quick conversation about this ‘make selection’ story?”
“Sure thing, be right over.”
Ron and Hector saunter up to Lori and Sally’s work area. “What’s up?”
“Sally and I have been working on this ‘make selection’ story, and we
think it might need to be broken up—refactored. Either that or we are mis-
thinking our objects. The story reads, ‘Allow the customer to make a selec-
tion from the available products, making sure that no expired products are
sold and that sufficient funds are available to pay for the product selected.’”
“We visualize this story going like this: 1) a selection object is created
when a customer inserts money or a debit/credit card; 2) the selection object
then monitors the console buttons for selection input; 3) when buttons are
pushed, the selection object puts them together as a kind of key or index value
and then asks the dispenser collection whether it has a dispenser matching
that key and whether it is operational (its product has not expired and it is not
empty); 4) the selection then asks the dispenser for the price of the product
and the accumulator for the amount of money available; 5) if we get all yeses,
we tell the dispenser to do its thing and give the customer the product. This
seems to capture the story, but it will involve a lot of potential error-handling
methods—in case the dispenser doesn’t exist or there are other problems.”
Hector looked a bit puzzled. “That seems to capture the story all right.
I don’t see any problem, but you are uncomfortable. Can you tell me why?”
 Chapter 4 Metaphor: Bridge to the Unfamiliar 105

Forward Thinking (continued)

“It just seems too complicated. The way we envisage building and
testing this, it isn’t simple enough.”
Ron said, “I think you have the story right, but you might want to
consider involving some different objects, and you might want to assume
that some of the things you are trying to do in this story are actually done
at a different point in time. Check me on this, Hector, to make sure that we
don’t invalidate your story.
“Your selection object is a good idea, but let’s simplify it. It will be
what Rebecca Wirfs-Brock and Alan McKean would call a structurer ste-
reotype object—that is, it will exist primarily to maintain a structural rela-
tionship: in this case, the parts of a selection. Some selections will have
only one part, as with the soda machine: a selection is just a numbered
button. For the candy machine, the parts are a letter and an integer. Your
idea of having it constitute itself in response to a money-inserted event
and button-pressed events is a good one.
“But we will introduce another object—a menu object. This will be a
collection kind of object, and it will represent, or proxy, the dispenser col-
lection. Visualize it like a two-dimensional table: column 1 is a key, and
column 2 is a variable.
“You remember variables? An object that contains an object reference
and a message to be sent to that object in order to change from an
unknown to a known value?”
“Yes.” Everyone agreed that they remembered what a variable object was.
“OK, now—assume that the menu is constantly updated whenever a
dispenser is activated or deactivated and that, at the time when a selection is
made, all you have to do is ask the menu whether it contains an entry at the
key value that matches your selection object. If it does not, a message is sent
to a userMessage object, which prompts the user for a different selection.
“If the selection is included in the menu, we involve another
object—a dispenseRule object. This is probably a singleton object, and it
lives in an instance variable of the vendingMachine object. Oh, and by the
way, your selection probably lives in another instance variable of the
vendingMachine while it is alive. The dispenseRule has registered with
the currentSelection variable (that is where your selection will live) to be
notified if the object-in-residence (your selection object again) generates a
selectionMade event. This is what will happen if the menu object confirms
that you are a valid selection.

(continued)
106 Object Thinking

Forward Thinking (continued)

“The dispense rule has two variables: one that sends a price message
to the product at the front of the dispenser, or maybe to the dispenser
itself, and a second one that sends the total message to the accumulator. It
also has one operator, and that is less-than (or greater-than, depending on
how you actually write your code to relate the two variables), and it eval-
uates to a Boolean true or false.
“If dispenseRule evaluates to true, the product is dispensed. If not,
another userMessage object is told to engage the customer in a dialog to
resolve the situation.
“Most of what I am suggesting is just moving responsibilities around
among objects—to simplify things for all concerned. But it does require
another user story—if Hector is agreeable—called ‘update menu.’ That
story will be executed whenever a dispenser changes its state.”
“No problem,” Hector says. “In fact, some kind of story of that type
would probably have been written anyway. The boss and I were discuss-
ing how we can build machines that contain different products in each
dispenser and have LEDs or icons or something appear on the face of the
vending machine to cue the customer about what is available inside—
your menu, in essence. I’ll write up the story, and we can bring it up at the
next game-planning session.”
“That help you two?” Ron asked Lori and Sally.
“Absolutely. We’ll have this done in no time with those simplifica-
tions. We’ll just stub in the menu behaviors for now until that story and
object are ready. Thanks, Ron, thanks, Hector.”

We have said that objects have access to all of the resources necessary to
do their jobs. In the case of a software object, this means that each object is
assumed to have access to all of the resources of an arbitrarily complex com-
puter system if necessary. This is one reason that Alan Kay calls objects “intel-
ligent virtual computers.” Because we are conceptualizing each of our objects
as possessing its own computer (virtual computer or thread), we have the foun-
dation for concurrent or parallel processing systems. Objects, like the people
we metaphorically equate them to, can work independently and concurrently
on large-scale tasks, requiring only general coordination. When we ask an
object collective to perform a task, it’s important that we avoid micromanage-
ment by imposing explicit control structures on those objects. You don’t like to
work for a boss who doesn’t trust you and allow you to do your job, so why
should your software objects put up with similar abuse?
 Chapter 4 Metaphor: Bridge to the Unfamiliar 107

Variations on the object-as-person metaphor include the object-as-agent

metaphor. The key difference between an agent and an object is autonomy.
Autonomy has at least two aspects: independence and freedom of action,
which both objects and agents share; and behavioral integrity, which an object
has but an agent does not. To understand what is meant by behavioral integrity,
consider the source of behaviors assigned to an object and the source of behav-
iors assigned to an agent. In the case of an object, the behaviors are intrinsic
to its nature. In the case of an agent, they are proxies of behaviors belonging to
the client of the agent. An agent does the things that the client is supposed to do
(would do if it had the time). This means that each agent is an idiosyncratic and
highly specific reflection of the needs of a client. An agent is not composable.
An object-thinking approach to agents would lead to refactoring the agent
into an intrinsic part (the behaviors that allow one to be a good agent) and a
variable part (the services that the agent agrees to perform on behalf of a client),
which we can call a contract object. Intrinsic behavior would include naviga-
tion abilities, knowledge of resources available, and general resourcefulness. A
contract is itself composed of different types of objects—goal objects, rule objects,
and results objects among them. Interactions between the agent object and its
current contract object would also require different script objects. The result is an
ensemble of objects working cooperatively instead of a single entity.
In the real world, we think nothing of people fulfilling multiple roles. People
are actors, and objects can also be actors. This metaphor leads into an entire
category of theatrical extensions, some of which are discussed in the next sec-
tion. The metaphor means that a single object can appear quite differently in
different contexts. Mel Gibson can be Hamlet or The Road Warrior, two dis-
tinctly different roles. A collection object can be used to replicate most of the
behaviors of a book or a parking lot, depending on the context and the values
and objects manipulated as it performs its services.
Many readers will be uncomfortable with the object-as-person metaphor.
A major source of discomfort arises from consideration of characteristics that
humans have that we obviously cannot replicate (and probably do not want to
replicate) in software. Emotions, true intelligence, and will are major examples.
The problem is made worse when descriptions of objects and their behaviors
seem to allude to precisely this kind of nonreplicable characteristic—for example,
“Objects are lazy.”
Effective use of metaphor requires constant awareness that a metaphor is
not a specification. It is often helpful to replace the metaphor—an object is a
person—with a supposition: what if an object were a person? You can then ask
questions about your design in this form: if an object were a person, would I
write my code this way? Here are two examples:
108 Object Thinking

■ If an object were a person, would I directly access part of its memory

without its knowledge instead of sending it a message asking for the
information I need? If a developer is obsessed with performance,
direct access is tempting. But reminding yourself that, as a person,
you would not like someone directly probing your brain without
your knowledge reminds you that this is a bad design choice. (It
leads to undesirable coupling.)
■ Technology exists that would allow me to make hardware connec-
tions to your brain. I could then build a control box that would allow
me to raise your hand whenever I pressed a button. Eventually this
same technology might allow me to make you perform a compli-
cated dance. Again, as a person, you might not like this, and neither
would an object if it were a person.

The object-as-person metaphor, and the parables you can tell using the meta-
phor, are simple, easy-to-remember proxies for heuristics and axioms of good design.

Software as Theater; Programmers as Directors

Engineering conveys an image of bridge or building construction. Software
engineering is therefore a metaphor for how software should be constructed,
that is, in a manner analogous to constructing a physical structure. Formalists
love this metaphor, but object advocates find it counterproductive. A better
metaphor for assembling objects to collectively perform tasks is theater.
Software development is analogous to casting and directing a play. The
first task is to select your players, the objects that will collectively complete the
expected tasks. Provide the cast with a script (cues and dialog). Test them
(practice or rehearsals) to make sure you have the right actors and the right
cues and dialog, and then, when you’re satisfied, put them onstage and raise
the curtain. If you have done your job well, the actors will proceed through the
play and the audience will be provided the service of entertainment.
Most software development involves the re-creation of an old standard—
an original play that was already cast and performed in the real world using
human and tangible objects (actors). Software developers face the same chal-
lenge as stage directors: how to make a play innovative and fresh without alien-
ating the audience by removing too much of the familiar and expected.
As fanciful as this metaphor may appear, it is deadly serious. And it reveals
one of the flaws of traditional software development: too much focus on the
artifact at the expense of the system in which the artifact will operate. If soft-
ware (computer) objects don’t simulate real-world entities in a reasonably
 Chapter 4 Metaphor: Bridge to the Unfamiliar 109

familiar manner, and if they are unable to interact with the other objects in the
real world in which they are to operate, they will fail. Software development is
reality construction (or reality reconstruction), just as Christiane Floyd and her
colleagues have asserted. And the theater metaphor helps us accomplish our
task by reminding us “that all the world is a stage” and that our artifacts are but
actors on that stage. (Chapter 7 of the Poppendiecks’ book on lean software
development explores this idea from a slightly different perspective.)

Note Occasionally developers have a chance to create a brand-new

reality. Most of what we get paid for, however, is simply reproducing
standard works. Sometimes that involves replacing a single actor, like
casting Mel Gibson as Hamlet instead of casting Laurence Olivier.
Sometimes it involves an almost complete surface change, like the
Star Trek episode that basically restages Moby Dick in outer space
(Picard as Ahab, the Borg as the white whale). The theater metaphor
adds the concept of verisimilitude (the appearance of being real) as a
criterion for good software and system design.

Sometimes the script followed by our object actors will be fixed. In soft-
ware, most batch processes would be considered to have a fixed script. But
more and more software needs a mix of fixed and extemporaneous scripts.
Extemporaneous scripts are those that are highly interactive and in which the
conversation is not predictable in advance.
Visual programming environments provide an illustration of this meta-
phor. Object (cast) selection is accomplished by dragging iconic representations
from a catalog to a workspace. Links between or among objects are established
by drawing lines to connect them with one another. Each link defines a circum-
stance in which one object communicates with another. (Events and messages
are two types of communication.) The collection of links established in the
workspace is the script for that group of objects.
The play (theatrical production) metaphor can be extended a bit further.
Plays come in many sizes, and the complexity of the script varies accordingly.
A one-person play is relatively simple to produce (but requires an exceptionally
talented actor), while a Cecil B. DeMille epic with a “cast of thousands” is con-
siderably harder to organize. The complexity of object-oriented application
software is in the scripting, not in the objects.
110 Object Thinking

Note Object-oriented programming, specifically Smalltalk, devel-

oped in parallel with graphical user interfaces. At Xerox PARC, they
were aspects of the same project. Early attempts to explain OO pro-
gramming tended to focus on applications having a graphical user
interface (GUI). Objects dramatically reduced the time and effort
required to create GUIs, which led to an inadvertent and unfortunate
belief that objects were primarily used for GUI design and develop-
ment. At the same time, an approach to program design, which the
Digitalk Smalltalk manuals called “cocktail napkin design,” gained
prominence. The idea was to draw your GUI, which would be easy and
obvious enough to draw on a cocktail napkin while you were discuss-
ing your program in a bar, and then design and build the model (the
objects behind the interface). GUI-driven design is at least as bad as
datacentric or algorithm-centric design. In each case, the developer is
taking one aspect of the implementation environment and using that
as the criterion for understanding and modeling objects. The typical
result is a program using a lot of dumb widgets that are observed and
updated by a monstrously complex behind-the-scenes controller.
Object thinking requires you to understand objects in terms of their
expected behaviors in a problem domain, never based on one or more
of their implementation aspects.

Our tools, as application software developers, for dealing with this com-
plexity are still quite limited. One need only reflect on the rapid accumulation
of visual clutter (overlapping lines, obscure icons) in a visual programming
workspace to see how limited our ability is to describe large-scale interactive
scripts. As with any other complicated task, we will attempt to solve this prob-
lem by decomposing our play script into act and scene scripts.
Four additional aspects of this metaphor deserve some discussion before
we move on. First, although we classify plays as being of various types depend-
ing on their complexity and scope, all theater is essentially the same, a group of
actors focused on accomplishing a particular objective, coordinated by a script.
With software, we distinguish between objects, components, applications, sub-
systems, and systems for our convenience. Close examination reveals that in
each case we have a number of objects focused on accomplishing a small list of
tasks while constrained by a guiding script. This tells us that we need to apply
the same principles of object philosophy whether we are constructing the most
 Chapter 4 Metaphor: Bridge to the Unfamiliar 111

specific class or a system with wide scope. We do not suddenly revert to old
habits just because the job is larger.
Second, replacing actors of similar talent and skill set should not require a
rewrite of the script for a play. Substituting objects capable of the same behavior
should not require a redesign of the software. We should be able, in fact, to
nuance or dramatically change the overall behavior of our software simply by
changing players. A drama can be turned into a comedy simply by replacing
dramatic actors with comedic actors. The latter receive the same cues and
deliver the same basic behavior (say the same lines) but use their innate abili-
ties to interpret the cues and respond in very different ways. This point will
become important later, when we discuss application frameworks.
Third, when the curtain rises, exceptional things (not planned for in the
script) will occur. It is up to the actors (objects) and their abilities, for the sake of
cooperative peer communication, to recover and keep things going forward.
Actors cue one another and do not have to rely on the director, nor does the
director need the complexity that would allow her to anticipate and control every
possible variation in a performance arising from unanticipated events.
Fourth, just as plays are categorized by genre, it’s appropriate to extend
the theater metaphor to software and classify systems into genres, based on
typical forms of organization, or architectures. An architecture is a patterned
way of organizing a set of actors, as is a genre. We generate expectations and
constraints that will apply to our actors based on the genre of the performance,
or the type of architecture. Architectures also provide general solutions or
frameworks that make it easier to conceptualize the organization of our cast.
Patterns (genres) provide “script templates” to which the designer adds detail
in order to construct the actual script used by an object collective to complete
its work.

Note Architectural patterns will be discussed in Chapter 9, “Objects

on Stage.” Pipes and filters, model-view-controller (MVC), black-
boards, and client/server are but a few examples of software architec-
tural patterns that will be discussed.

Care must be taken, however, to make sure that assumptions implicit in

the architecture (genre) do not contradict or contravene object thinking. It
would be difficult, for example, to use the typical hierarchical control script
that’s embodied in the infamous program structure chart popularized by Yourdon
and Page-Jones in an object fashion.
112 Object Thinking

Ants, Not Autocrats

Hierarchical and centralized control is anathema in the object paradigm. It is
replaced with a kind of blind coordination, as exemplified in the traffic signal
example in Chapter 3, “From Philosophy to Culture.” A traffic signal is blind in
the sense that it does not need any awareness of other objects or their goals to
accomplish its own tasks. Any sense of traffic control has been distributed to
the collection of objects in the intersection and not to any single object.
The traffic signal assumes responsibility for monitoring the passage of
time and cycling through a change of states at appropriate intervals (green for
20 seconds, yellow for 10 seconds, red for 30 minutes).2 It also assumes respon-
sibility for notifying others of its current state by broadcasting that state via an
externally observable colored light. Automobile (or driver) objects assume
responsibility for inhibiting or expressing their own behavior (stop on red, go
on green, accelerate on yellow)3 as a consequence of their awareness of the
signal. Neither automobiles nor drivers know anything about the workings of
traffic signals, just as signals know nothing about automobiles or drivers.
This kind of blind coordination seems to work well in small-scale examples
such as the traffic signal, but does it scale up? The answer, suggested by ants,
termites, and biological communities, is yes. “Hive communities” collectively
construct extremely elaborate structures and efficiently exploit natural
resources (such as food) without the need for architects or overseers. No single
ant is in charge of making sure that a group of ants perform. Food foraging
begins when a single discoverer ant broadcasts the discovery of food to the
other ants by exuding a particular pheromone. Other ants detect the phero-
mone and respond by moving to the food source and then back to the hive,
also exuding the same pheromone. No ant is aware of the identity of
any other ant. They do not seem to care whether other ants are around. They
simply detect an event (receive a message) and respond according to their
intrinsic nature.
The attempt to establish centralized economies and management in the
precollapse Soviet Union is a contrasting example of autocratic top-down control.
It suggests that although it might be possible to construct very small-scale,
control-oriented systems, it does not work on a large scale. The emerging dis-
cipline of complexity theory also provides insights into the limitations of control
in large systems.

2. OK, only subjectively.

3. Observed but highly improper behavior.
 Chapter 4 Metaphor: Bridge to the Unfamiliar 113

Note If the idea of patterning software architectures on ant or termite

colonies makes you uneasy, you might consider the work of Marvin
Minsky (a significant contributor to OO programming as well as to AI the-
ory) and the work of Murray Gell-Mann, Stuart Kaufmann, and many oth-
ers at the Santa Fe Institute. In his Society of Mind book, Minsky posits
the construction of intelligence by utilizing nonintelligent actor/agents
(objects) that make simple decisions based on their own local awareness
of themselves and their individual circumstances. These simple agents are
aggregated in various ways until they form a “society of mind,” which is
also the title of the book in which these ideas are formulated. Gell-Mann
and Kaufmann are among the founders of the Santa Fe Institute, the
center for the study of complex adaptive systems. Complex systems are
characterized by simple elements, acting on local knowledge with local
rules, giving rise to complicated patterned behavior. And of course, you
should consider the fact that XP asserts that large systems can, do, and
will emerge from the creation and implementation of small local systems.

Note In human systems, the presence of strong central control has a

debilitating effect. Humans lose their capability of independent action
and become dependent on the presence of the controller. If that control-
ler is removed—as it was in the former Soviet Union—it can take a long
time for the “controlled” to regain their ability to function autonomously
and cooperatively. Software objects designed to function in response to
a master controller cannot function in the absence of that controller.
This is the antithesis of design based on object thinking.

These two counterexamples provide the basis for the metaphor that
objects are coordinated as if they were ants and that no object attempts to
assume the role of autocrat controlling the behavior of other objects.

Two Human-Derived Metaphors

Two metaphors, inheritance and responsibility, arise from our observation of
human beings. Both are important, and both are subject to some degree of con-
fusion because the terms are used somewhat differently in different contexts.
114 Object Thinking

Inheritance
Humans naturally aggregate similar things into sets (or classes). Another “natural”
kind of thinking is to create taxonomies—hierarchical relationships among
the sets. The most common example of this kind of thinking is Carolus Linnaeus’s
taxonomy of flora and fauna and the subsets of that taxonomy that are general
common knowledge. Fido is a dog (example of aggregation and the subsequent
identification of a set). The set Dog is a subset of Canine, which is a subset of
Mammal, which is a subset of Animal. (The example taxonomy isn’t complete,
nor is it intended to be accurate—merely illustrative.) We could also say that a
dog is a kind of canine, which is a kind of mammal, which is a kind of animal.
Taxonomies are tree structures.
Another kind of tree structure—one that actually employs the term—is a
genealogical chart, a family tree. Because both the taxonomy and the geneal-
ogy chart use the same structure, a hierarchical tree, they have become a kind
of conflated metaphor.
The terms parent and child, for example, are clearly appropriate for gene-
alogy but are somewhat suspect when applied in the context of a Linnaean
hierarchy. Nevertheless, it is common to speak of the superset/subset relation-
ship in terms of parents and children. A superclass is Parent, and a subclass is
Child. From here, it’s but a short step to talk about children “inheriting” from
parents.
At this point, the metaphor can be helpful or potentially misleading
depending on how it is used.
According to the presuppositions of the object paradigm, a child class is a
behavioral extension of the parent. A dog has all the behavior of a mammal
plus some additional behavior that’s specific to dogs. If we say that a dog inherits
the behaviors of a mammal and mean by that statement that a dog can be asked
to do anything that a mammal can do, we are using the metaphor properly.
Too often, however, the metaphor is used to assert that the child class
inherits the internals of the parent class, an allusion to the fact that biological
organisms inherit the DNA structures of their parents. This is a poor and poten-
tially misleading use of the metaphor.
Behavior is the abstraction that we use to differentiate among objects, and
it should be the only criterion that we use to establish our taxonomy. Using any
other criteria will make our taxonomy more complicated at a minimum and
erroneous at worst. In the real world, errors such as racism and sexism can be
seen as characteristic-based rather than behavior-based taxonomies—with the
obvious negative consequences.
In the context of software objects, creating taxonomies based on internal
structure (attributes or operations, for example) causes problems in numerous
 Chapter 4 Metaphor: Bridge to the Unfamiliar 115

ways. One example is the need to create new classes of objects such as Customer
and CreditCustomer just because one has different characteristics in some con-
texts than in others (a credit rating, perhaps). It can also lead to an apparent need
for multiple lines of “inheritance” when an object has characteristics that are part
of the structure of two or more potential parent objects.
The desire for a child class to inherit internals of its parent classes can be
better accommodated if we change the notion of inheritance from DNA to
assets. It has been noted that an object has access to whatever resources it
needs to fulfill its behavioral expectations. If we say that child classes have
access to the resources of their parents via inheritance, that is, by virtue of the
parent-child relationship, our use of inheritance remains consistent with the
object paradigm.

Responsibility
We know an object by what it does, by what services it can provide. That is to
say, we know objects by their behaviors. We are not interested, of course, in
just any old behavior. We have specific expectations of our objects and are sur-
prised if they deviate from those expectations. Because we expect objects to
exhibit certain specific behaviors, we tend to hold them accountable for those
behaviors. (We experience negative reactions such as disappointment and even
anger if the objects “let us down.”) We expect them to “perform as advertised.”
Our thinking about the abilities of objects shifts from simple awareness of
behavioral possibilities to expectations of service, of responsibilities. The capa-
bility is unchanged, but our perception of that capability is different enough to
merit the use of responsibility as a label.
If an object states that it is capable of providing a given service, it should
perform that service in all circumstances, and the results should be consistent.
If an integer object says, “I can add myself to any integer you provide and
return to you the resulting integer,” it would be very irresponsible if sometimes
it returned a floating-point number or, worse, if it were an integer that was
other than the one representing the summing process.
Responsibility implies that an object must assume control of itself. It must
be capable of assuming responsibility for its own maintenance, for notifying
others of any interesting and shareable state changes that it might experience
and that other objects might need to be aware of, for making sure it’s persistent
when it needs to be, for protecting its own integrity, and for responding appro-
priately to requests for service from multiple clients. Just as it’s improper for one
object to assume a role as controller or manipulator of another, it’s improper for
an object to fail to assume responsibilities that make the need for external
control unnecessary.
116 Object Thinking

The notion of self-responsibility will play an important role in how we

decide to allocate behaviors across a collection of objects.

Thinking Like an Object

Metaphors such as those just introduced should shape our thinking about objects
and about software constructed from those objects. Internalizing these metaphors
and the philosophical presuppositions stated in earlier chapters allows you to
start “thinking like an object.” Until you are able to do so, and more important,
until you can extend your thinking into new areas, you will fail to realize the full
potential of objects. In large part, the internalization process will occur over time
and with experience. It is possible to adopt certain techniques and use certain
models to enforce the practice while you are gaining experience. Later portions
of this book will explore and discuss some of those models.
Most of the models discussed will bear a striking surface similarity to models
constructed using classical software modeling techniques. Syntactically, the
variation between many objects and classical models is minimal. It’s the semantic
content of the models—derived from how we accomplish our analysis and
decomposition of our domain—that will vary.
Because the models are so similar in syntactic structure, it will be very easy
to revert to old habits, to use the new models in old ways. A partial defense
against this tendency is the adoption of a new vocabulary. In the next chapter, we
will examine a vocabulary that reflects both the philosophy introduced earlier
and that is consistent with the metaphors presented in this chapter. Whenever
the vocabulary definitions seem strange or unduly restrictive, remember the
philosophy and the metaphors that inspired them.
Vocabulary: Words
to Think With
Language shapes our ability to think. Words provide us with representations of
particular thoughts and offer a mechanism for communicating thoughts among
ourselves. The richness, variability, and subtlety of our thoughts are reflected in
the size of our vocabulary. When we discover new things, we invent (or borrow)
new words to express our discovery. When we want to make critical distinctions
between or among similar notions, we use specialized vocabularies. The use of
special vocabularies is common in our professional lives because, as specialists,
we need to become familiar with unique ideas and to make important distinctions
that as laypersons we might not make.
Object-oriented development has its own specialized vocabulary for all of
the aforementioned reasons. The unique vocabulary of objects has another
important purpose: to differentiate object concepts from traditional software
concepts that on the surface appear to be similar. Object vocabulary is deliber-
ately different and is focused on communicating object thinking. To ignore the
vocabulary or to trivialize the differences between object and traditional terms
is to make a significant error.
Perhaps the most asked question in an introductory object class is some
variation of, “What is the difference between a (class, method, message) and a
(module, function, procedure call)?”
In one sense, this question can be answered only with, “Nothing, there is
no difference.” Given this answer, the next question becomes, “Why then do
you use these silly new words?” Underlying this question, of course, is the
sneaking suspicion that the object advocate is a snake oil salesman trying to
confuse the innocent customer with a fancy polysyllabic vocabulary.

117
118 Object Thinking

Forward Thinking
It’s Not a Code “Smell,” But …
“Every time I look at this code, it seems strange—but not strange,” Jon
comments to Samantha. “It’s pretty ordinary for the most part, but I have
this constant nagging feeling that something is different—not wrong, just
different.”
“Not surprising,” replies Samantha. “This is probably your first real
object-oriented project as well as your first XP project.”
“What do you mean? I’ve been writing Java for several years; doesn’t
that count as OO?”
“Not necessarily, because what you code is dependent on how you
have thought about the problem space and how you have decomposed
that space into objects. Just using a particular language doesn’t mean you
build objects; only object thinking about decomposition assures that you
will build objects.”
“Huh?”
“Let me show you. Remember the ‘Serene Meditation’ story we
worked on last week?”
“Yeah, our customer has some imagination when he writes story
titles. ‘Serene Meditation: The machine goes into a power-save mode until
a potential customer triggers an induction field around the machine or
calls the machine’s published phone number. When its meditation is inter-
rupted, the machine lights up, its selection menu is presented, and the
money-accepting mechanisms are initiated.’”
“Yeah, he does, but when we first wrote the code for that story you
had two ‘controller’ objects, phone and inductionField, that tried to keep
track of the other components and send them wake-up messages. When I
pointed out the duplicated code, you quickly refactored into a single con-
troller, but then you had to include a case statement to notify different
objects depending on the actual source of the wake-up event. Only after
we talked about MVC and using event dispatchers and event registration
did we hit on the correct solution: let the machine, selectionMenu, coin-
Accepter, billAccepter, and cardAccepter objects register with the induction-
Field or phone object so that it’s notified if it detects the unique event it’s
looking for. When we wrote the code, it didn’t look radically different, but
the organization of the code to form the methods was different.

(continued)
 Chapter 5 Vocabulary: Words to Think With 119

Forward Thinking (continued)

“And last week, we were working on the form object and had really
different ideas of what its methods should be?”
“Yeah, I had a whole bunch of methods, mostly setters and getters
for attributes. You had only three or four methods that you thought nec-
essary. Turns out I was thinking about attributes all wrong.”
“Not all wrong, and according to some object theorists not wrong at
all. When Coad and Yourdon published their first book on OO analysis,
they defined objects and attributes as if they were standard data entities.
Schlaer and Mellor, Rumbaugh, and a host of others did essentially the
same thing. UML, which you used a lot before coming here, allows greater
latitude in the definition of an attribute, but the base definition is still ‘a
named characteristic with a value.’ You were still thinking of attributes as
some characteristic of an object, the value of which the system had to
remember.
“So you were thinking of a form in terms of a lot of attributes: form-
Number, dateCreated, lastRevisionDate, title, and so on. I was defining a
form as an object that contained other objects, a collection, and an object
that could identify itself, display itself (with help from the objects it con-
tained), and describe itself.”
“Yeah, all the attributes I was listing ended up being inside a descrip-
tion object—another kind of collection, kind of like a dictionary.”
“So you see, it’s not the code that’s different—it’s the objects and the
distribution of methods across the set of objects that’s different.”
“And the vocabulary is different too. I have to keep reminding myself
that things such as functions and methods look essentially the same, but
they derive from very different ideas. A method reflects a behavior expected
in the domain, and a function reflects a unit of execution in the machine.”
“Got it! Now back to this ‘reorder product’ story.”

The object difference is not readily apparent at the level of implementation.

As noted earlier, once software is in the machine it’s “all ones and zeros.” The
difference is how we think about the problems and solutions. This difference is
most manifest in how we do decomposition and how we decide to distribute
responsibilities across a community of objects. A message and a function call
might be identical, in syntax and in how they are implemented inside a machine.
But who sends the message to whom, how that message is interpreted, and what
kind of code will be necessary to respond will be quite different.
120 Object Thinking

Object vocabulary is also different because of the underlying philosophy

that motivates thinking in new terms and because we are employing different
metaphors for what we are about and what we are building. Different philos-
ophy + new metaphors = object vocabulary.

Important Object vocabulary is first and foremost a technique to help

developers avoid the mistake of thinking about solutions using old
mental habits.

Although numerous terms are employed in the discussion of objects, not

all of them have the same degree of importance. Relatively few terms both need
to be understood as fundamental and embody the majority of object philoso-
phy. Other terms extend or elaborate aspects of essential terms and allow us to
say other interesting things about objects or the way software objects might be
implemented. Still another set of terms applies almost exclusively to software
objects; these terms are a function of the implementation language selected for
development. A fourth set consists of auxiliary terms. Table 5-1 lists the vocab-
ulary terms in the four categories: essential, extension, implementation, and
auxiliary.

Table 5-1 Categorized Object Vocabulary

Essential Implementation
Extension Terms Auxiliary Terms
Terms Terms

Object Collaboration/ Method Domain

collaborator
Responsibility Variable Business requirement
Class
Message Dynamic binding Business process
Class hierarchy reengineering (BPR)
Protocol
Abstract/concrete Application
Inheritance
Delegation
Polymorphism
Encapsulation
Component
Framework
Pattern
 Chapter 5 Vocabulary: Words to Think With 121

Essential Terms
The following terms have the greatest philosophical importance—they embody
most of the philosophy that makes the object difference.

Object
Objects are the fundamental units of understanding. We define the world (and
the world of software) in terms of objects. Everything is an object! An object is
anything capable of providing a limited set of useful services. Metaphorically,
an object is like a human agent or actor. We decompose the complex world
around us in terms of objects, and we assemble (compose) objects in various
ways so that they can perform useful tasks on our behalf.
Our understanding of an object is, and should be, based on that object’s
public appearance: its “family name” (class) and an “advertised list of services”
(protocol) that it is willing to provide. Every object has a “personal name,” which
is a unique identifier, and a “family name,” which identifies the set (class) of sim-
ilar objects to which the particular object belongs. Most often we speak of the
family (or class) name of the object. That name is descriptive of the object and
should convey a general sense of the kind of services it might be able to provide.
The set of advertised services usually takes the form of a list of syntactic phrases
that we can use to invoke each behavior. Separately documented is a list of
potential states that the object might be in. Not all object states are included in
this list, only those that the object wants to share and those of which other objects
need to be aware. Together these lists constitute our behavioral promise to the
world at large. We (speaking as an object) do not share with the world any notion
of our internal structure and certainly no sense of how we do what we do.
Every object has access to whatever knowledge is required to perform its
advertised services. This does not mean that the object contains that informa-
tion. Information can also be accessed by asking another object, supplied as
part of a request, or calculated upon demand.
Traditionally, software was conceived in terms of passive data and active
procedures. One of the more famous definitions of a program is “algorithms plus
data structures.” This is also called the pigeonhole approach to software—passive
bits of data sitting in boxes (such as those seen in a post office sorting room,
which are called pigeonholes) where procedures come to remove a bit of data,
transform or use it, and then put it back for the next procedure to access.
Because everything is an object, there is no data. One of the principal
ideas of traditional software approaches has disappeared. Everything, including
characters and integers, is an object responsible for providing specific services.
The elimination of passive data has important consequences for object design,
a topic that will be discussed in depth later in this chapter. The consequences
122 Object Thinking

manifest themselves at the level of design and are felt minimally at the level of
code. This should be reassuring to those (a vast majority) who are using typed
languages such as Java. When you actually write code and are compelled to
declare types for variables or use objects such as strings and characters and
numbers solely for their ability to represent data, you won’t be violating any
principles of object thinking. You will find that you have fewer objects repre-
senting passive data and are much more likely to create an abstract data type to
represent a data object instead of using constructs built into the language.
Therefore, the three popular graphic models of an object shown in Figure 5-1
are misleading and should not be relied upon. The donut model (what Ken Auer
calls the soccer ball model because the dividing lines resemble the seams of a soc-
cer ball) perpetuates the classic ideas of passive data and active procedures and
better describes a COBOL program than an object. The animated data entity
model also perpetuates the outdated separation of data and procedures and com-
pounds that error by focusing on the distribution of data items—attributes—across
a group of objects. Procedures are appended to the objects containing the distrib-
uted attributes. This data-driven approach to object modeling is quite popular but
is not consistent with object thinking.

Encapsulation
Barrier
Class Name
operationX
Attribute List
Attributes are
Data typically determined
operationY according to the rules of
Structure
data modeling
Operations List

operationZ Operations are defined

as functions operating
on the data attributes
listed above

Class Name
Responsiblity Collaborator
a short list of Occassionally the satisfaction
behaviors appears of a responsibility requires
here. services from another object -
short = ??? a collaborator. If so the class
name of the collaborator
appears here.

Figure 5-1 Three popular graphical object models: “soccer ball,”

F05MQ01

animated data entity, and CRC card (left to right).

 Chapter 5 Vocabulary: Words to Think With 123

Another, and essential, way that objects differ from the modules at the
heart of traditional software decomposition is that an object represents some-
thing naturally occurring in the problem domain, while a module represents a
logical aggregation of elements appearing in the solution space. This is another
reason why the foregoing graphical depictions are misleading and should be
avoided. All three diagrams depict what is to be built instead of what is to
be modeled. Object discovery and specification must be domain driven!

Note What of middleware, networking protocols, persistence mecha-

nisms (databases), and legacy systems? Are these part of the problem
space, the domain, or part of the implementation space? The answer,
as unhelpful as it might be, is that they can be either. You discover which
they are by initially thinking of them as another domain object. Name
them, and assign responsibilities to them. As you continue with your
development, see whether the objects retain those responsibilities or
whether all of the behaviors are reassigned to other objects. If the
objects retain responsibilities, they are part of the problem space; if not,
they are simply another solution space mechanism. For example, you
might have a database object with the responsibilities of enabling object
persistence, scheduling access to stored objects, enforcing object integ-
rity constraints, providing backup services, and managing update trans-
actions. It’s quite conceivable that all of these responsibilities could
be transferred to other objects, leaving the database without a role.
(See Chapter 10 for further discussion of objects and persistence.)

Responsibility
Responsibility means a service that an object has agreed (or been assigned) to
perform. Objects are charged with performing specific tasks. Each task is a
responsibility. The term responsibility is used to aid us in discovering who (which
object) should be charged with a task without the need to think about the object’s
structure. (We should not know its structure.) Responsibilities are characterized
from the point of view of a potential client of a service. I ask, “Who would I rea-
sonably ask for this service?” and my answer determines which object becomes
responsible for satisfying that type of service request.
124 Object Thinking

Note XP practitioners talk of test-driven development and story-driven

development. Both are instances of domain-driven development
because both assume that tests and stories reflect the domain and the
input of the on-site customer. There is an aspect to testing that is not
directly reflective of the domain since some tests are directed to deter-
mining whether “it was done right” as well as “it does what is wanted.”
The former set of tests determines the congruency of a particular solu-
tion with the dictates and constraints of the solution space.

The usual definition of an attribute is a characteristic of an entity, the value

of which must be remembered by the system. In a data-driven approach, the
attributes of an object are discovered first and then responsibilities are meted
out as a function of which object holds which data. A behavioral approach
mandates the assignment of responsibilities first. Only when you are satisfied
with the distribution of responsibilities among your objects are you ready
to make a decision about what they need to know to fulfill those responsibil-
ities and which parts of that knowledge they need to keep as part of their
structure—in instance variables or attributes. This is the single biggest difference
in definition between data-driven and behavior-driven or responsibility-driven
approaches to objects.
Responsibilities are not functions, although there is a superficial resem-
blance. Perhaps the easiest way to differentiate between a responsibility and
a function is to remember that the former reflects expectations in the
domain—the problem space—while the latter reflects an implementation
detail in the solution space—the computer program. Recall the applicability
graph, Figure 3-4, and its division of the world of software into two realms.
Use responsibilities when discussing anything in the realm defined by the
upper left half of the graph and function for things far to the left and bottom
of the graph.
An object might perform any of four basic types of services. (For a more
detailed exploration of types of services an object might perform, see Rebecca
Wirfs-Brock and Allan McKean’s book Object Design: Roles, Responsibilities,
and Collaborations.)

Maintain and supply on request one or more units of information.

The information (everything being an object) is an object such as a string or a
character or a number, the value of which conveys meaning to an observer or
a user. A person object might agree to provide its identification, which means
 Chapter 5 Vocabulary: Words to Think With 125

it agrees to provide you with a string, the value of which you would recognize
as a unique identifier.
A more complex example would be an object, a product perhaps, that
agrees to provide you with its description. In response to your request, this time
the object gives you a description object. A description object can be thought of
as a collection of labels and associated values called a value holder (essentially
the same thing as a Dictionary in Smalltalk or a Map in Java). (See Figure 5-2.)
You then ask the description object for a particular value associated with a
specific label.

Figure 5-2 A value holder, a simple two-dimensional array with the con-
F05MQ02

tents of the first column restricted to labels and the contents of the second
restricted to data objects such as strings, numbers, dates, and characters.

Perform a computational task.

This is a straightforward responsibility as long as you remember that computations
must be performed by an object on itself. A number object, for example, can add
itself to another number,1 but it would be inappropriate for a calculator object
to add two numbers together. A calculator is a kind of control object and should
be avoided, both because it violates the object paradigm and because it would
have to be a very large and complicated object, another trait to be avoided.

1. In this example, the second number does assume a kind of passive role, being subsumed by the active
number doing the calculation. But either number could have been the active one, so the principle of
objects acting only on themselves is preserved.
126 Object Thinking

Note A calculator object might still exist in your solution—but as a

façade for the community of objects doing the calculating rather than
the single object doing the math. Façades are quite useful because
they provide a single point of reference for other objects seeking ser-
vices. They can also provide a simple set of services, all or most of
which are parsed and passed to objects behind the façade for actual
implementation.

This notion can be generalized so that a mathematical expression, which

is a specific kind of computation, should be responsible for evaluating or solving
itself. An abstraction of this service is a class of objects named SelfEvaluatingRule.
This class will be discussed in detail in Chapter 10.

Report on or update the state of the object.

Sometimes an object will change its internal nature (its state), and on those
occasions it might be expected to report that change to others. For example, a
seat object might have the state of empty, reserved, or occupied. When it changes
from one state to another, potentially many other objects—potential users of the
seat—would need to be notified of the change in state. The object should not,
however, know anything about potential clients and should only indirectly keep
a list of those objects that might want to be notified of state changes.
Every object should have the ability to accept registrations for event noti-
fication, even if many will never take advantage of that capability. Each regis-
tration is, of course, itself an object. Registration objects are accepted by a
Dispatcher object, as illustrated in Figure 5-3. A registration message (please
register me for event X) would be sent by an object interested in one of the
states you might find yourself in and of which you have indicated a willingness
to notify others. The potential client of an event notification will send your
eventDispatcher object a registration object (see Figure 5-4) and ask your event-
Dispatcher to add that object to the queue associated with the event of interest.
The registration has an internal structure consisting of the name of the object
sending the registration and the message that the client object wants sent to it
when the event occurs. The dispatcherObject stores the registration objects it
receives in notification queues associated with each potential event. When the
state change occurs (an event is generated), each registration object in the
appropriate queue is told to send its message to its recipient.
 Chapter 5 Vocabulary: Words to Think With 127

Event Dispatcher

Event Registration Queue

Evnt1 [object message] [object message]

Registration
Evnt2 Queues can be
FIFO or prioritized
Evnt3

Figure 5-3 A Dispatcher object. Events to be dispatched are located in

F05MQ03

the first column, whereas the second column is a queue of event registra-
tion requests.

[object message]
object is the name or identity of an object which is sent
the message when the event associated with the queue
is detected. An optional priority could be added to the
dyad if the event queue allowed priority dispatch
instead of first in first out.

Figure 5-4 A registration request consists of a simple dyad—who is to

F05MQ04

be notified (an object) and the means of notification (a message).

Coordinate other objects.

The key here is to maintain the blind coordination principle. Examples might be
a dispatcher that routes messages without any knowledge of why the messages
are being sent or a queue that coordinates the sending of a group of messages
and hence the activation of the objects receiving those messages but does so
only because its nature is to maintain order among a collection of objects and
release them from the queue at appropriate times.
The same dispatching objects (Dispatcher and Registration) are used to
effect the blind coordination desired. Individual objects can coordinate with
one another by exchanging registrations, and group behavior can be effected
by allowing a dispatcher to have “global” visibility—again, like the traffic signal.
The traffic signal example lacks any obvious event registration and dis-
patching, but they are there nevertheless. “Registration” occurs when the driver
of a vehicle or a pedestrian at the crosswalk gazes at the traffic light or walk
signal. “Dispatching” occurs when the light from the signal is broadcast to all
who can see.
128 Object Thinking

Message
A message is a formal communication sent by one object to another requesting a
service. A message can be imperative, informational, or interrogatory in nature.
An imperative message directs an object to make some change to itself. No
response is expected or required. The object is assumed to have made the appro-
priate change. (It’s quite possible, of course, for an object to offer confirmation of
an imperative message by returning some specific kind of object, but care should
be taken that this capability not be misused to create controller objects.)
An informational message is similar to an imperative in the sense that no
response is expected. It differs from an imperative in that there is also no expec-
tation that the receiving object will do anything at all in response to the message.
An interrogatory message is any request for a service. The object always
returns an object (a typed value or a signal/interrupt in most popular program-
ming languages) that encapsulates (embodies) the result requested. The
returned object can be simple (a character or a string) or arbitrarily complex.
Messages frequently take the following form:
Receiver Selector (Arguments) ← returnedObject or Receiver.Selector
(Arguments):ReturnedObject.

where
■ Receiver identifies who is being sent the message. It might be a spe-
cific named object (the object Sara), a generic object (aPerson), or
the name of a place where an object resides (a variable). In the case
of a variable, the object in residence actually receives and responds
to the message.
■ Selector identifies the semantics of the message, the essence of the
request. Selectors can be simple symbols (such as the mathematical
operators, +, *, /, and so forth) or descriptive phrases (nextObject-
InLinePlease).
■ (Arguments) are objects sent along with the message and are optional.
They’re optional in the sense that some selectors do not require an
object, but if a message signature indicates that an argument is
required, it is mandatory. Arguments are used when the receiver of a
request for service also expects the requester to provide some of the
information needed to perform that service.
■ returnedObject is an arbitrarily complex object containing or repre-
senting the result created as a consequence of receiving the message.
In the case of imperative and declarative messages, the object
returned is “self,” that is, the object that received the message is indi-
cating it’s still available to you (usually by keeping an active pointer).
 Chapter 5 Vocabulary: Words to Think With 129

In Java, and most popular languages, imperative and declarative

messages are void functions. No return is expected. If the object
changed in any way as a result of receiving the message, the “self”
that is returned is the newly constituted “self.”
What happens when you send a message and get back the wrong object,
do not get a response at all, or get an explicit error message instead of the
expected object? You figure out what went wrong, determine how the object
should behave in such situations, write tests for that behavior, and code a new
version of your object. The real question, however, is, “Who handles errors?”
The answer is to make error handling an object responsibility—one shared by
every object—and to let each object respond to those error conditions that it
can resolve by using the knowledge it already has in its possession regarding
itself and what it is trying to accomplish.

Interface (Protocol)
The collection of messages that an object responds to and the state changes it
indicates it will accept registrations for constitute its interface. The term proto-
col is usually reserved for that portion of the interface listing the messages that
the object is willing to respond to. A protocol might include messages of the
sort, “What state are you in?” or “Are you in state X?” but those messages should
not be confused with the innate ability of every object to notify others of a
change in its state. The list of states and the mechanism for event registration
and dispatching are quite separate from the message protocol. The syntax of
each message in the protocol is specified and is considered the message signa-
ture. Where arguments are expected along with the message, the signature spec-
ifies the class (or type in some languages) of the object expected. The class (or
type) of object returned in response to the message is also part of the signature.
The preceding terms are considered essential because they embody every-
thing you need to know about objects to successfully decompose a domain,
identify and distribute responsibilities among a collection of objects, create a
taxonomy of objects, and even create scripts to guide objects in the completion
of specific tasks.
These terms do not provide sufficient definition for those charged with
actually implementing software objects. They provide a specification only.
Implementation requires concepts associated with the solution space—a pro-
gramming language and an implementation platform. Also required is vocabu-
lary for describing the “internals” of objects. A supporting vocabulary addresses
these needs.
130 Object Thinking

Extension Terms
The following terms, although of less philosophical importance than the essen-
tial terms, are nonetheless commonly encountered. They are useful because
they nuance or extend aspects of the essential terms and because they intro-
duce some ideas necessary to thinking about implementation.

Collaboration and Collaborator

Collaboration and collaborator are terms referring to a particular type of object
cooperation and the object relied upon for that cooperation. Objects are social
and convivial things, constantly exchanging messages with one another and
cooperating to complete tasks beyond the capabilities of any single object.
Collaboration is a form of cooperation that is treated a bit differently from all
the other forms of cooperation.
Collaboration occurs when
1. Object A receives a request for one of its advertised services.
2. While in the process of satisfying that request, it needs to ask for a
service from object B.
3. Object B is not an object occupying one of object A’s instance vari-
ables, a temporary variable declared in the method that object A is
executing in order to satisfy the original request, or an object sup-
plied to object A as an argument of the message requesting the
service—that is, it is not an object currently residing inside object A’s
encapsulation barrier. This definition of collaboration correlates
nicely with the Law of Demeter—a general style rule (proposed by
Ian Holland in 1987) for object-oriented systems.
Object B becomes the collaborator, a covert assistant to object A.
The exchange between object A and object B occurs inside the encapsu-
lation barrier of the object. It’s this covert aspect that makes this particular
exchange different from all other object-object messaging. Collaboration is an
aspect of how an object satisfies a request. Collaborations always involve some
degree of coupling between both parties of the collaboration, so the number of
collaborations should be minimized to the greatest extent possible.

Class
Class is a term with many meanings, including

■ a label for a set of similar objects

■ an exemplar object
 Chapter 5 Vocabulary: Words to Think With 131

■ a storage location for knowledge (rare) or behavior mechanisms that

are identical for all members of the class
■ an object factory

Following is an in-depth explanation of the foregoing meanings.

■ Class as set. An application of the principle of aggregation, noted in
Chapter 4, a class is a convenient label for a group of objects that have
the same behaviors. Instead of referring to Alice, Dick, Jane, and Bob,
we can refer to Person, meaning any one of those individuals. Objects
that are included in the class set are said to be instances of that class.
■ Class as exemplar object. A confusing (for the beginner) but
common practice is to use the terms class and object interchange-
ably, with the context indicating which is meant. For example, we
talk about modeling an object’s behavior, and we might even visual-
ize a particular object (such as Mary), but we talk about the class,
saying, “A Person can identify itself.” When we do this, we are using
the class as an exemplar of the objects it instantiates.
■ Class as storage location. This use of the term is relevant only in
the context of software objects. The mechanisms that allow an object
to fulfill its responsibilities are, in the case of software objects, blocks
of implementation language code. It’s more efficient, both in terms
of physical storage and, more important, in terms of maintenance, to
store such a mechanism in one place: in the class. Instances of the
class are given access to the common mechanism when required. In
the real world, this use of class does not normally occur because
storage and maintenance are not usually important issues. All human
beings, for example, have bits of neural tissue that, when activated,
allow them to respond to the message, “What is your name?” The fact
that this bit of tissue is duplicated in every human being does not
bother us. It is unlikely that we will ever find a more efficient bit of
neural tissue and therefore unlikely that we will ever recall all human
beings for a “brain update.” Conversely, in the case of software we
do not want to duplicate even one line of code several billion times,
and we are quite often faced with the need to update an algorithm or
a bit of code. Therefore, it makes sense to store that code (mecha-
nism) in one place. It is also possible to store information shared by
all instances of a class in the class itself, and for the same reasons.
This use of a class for information storage is rare because the value
of the object stored in the class variable must be appropriate for all
instances existing at a particular point in time, a condition that is
seldom possible to satisfy.
132 Object Thinking

■ Class as object factory. This is another metaphor, which states

that a class has the responsibility of creating new objects, new
instances of the class. To do this, it needs a specification for an
instance (an object). When we define a class, we commingle what
are really two definitions, one for the class itself and one for the
objects it will be responsible for creating. This will be clarified later
in the definition of implementation terms.

Class Hierarchy (Library)

A class hierarchy, or class library, is the taxonomic organization of a group of
classes. In some implementation languages (Smalltalk), this is an organized
hierarchy, while in others (C++), it might be a simple shared library with no
hierarchy except compilation dependencies. A class hierarchy should, ideally,
function as a kind of index to a potentially large group of objects. For example,
we might be looking for an object that collects things and go to the class hier-
archy to find collection classes. Upon further reflection, we might remember
that things in the collection must be maintained in a particular order and find
the subclasses OrderedCollection and SortedCollection. This process could con-
tinue until we discovered the class we needed or found that none was avail-
able, in which case we would create the class and add it to the taxonomy
(usually at the place where our search ended).
The organization of a class hierarchy—the is-a-kind-of relationship—does
not provide the only index to the classes and methods in the library. Good
browser tools provide many other indexes, including senders of message, imple-
menters of message, category of class, and category of method, among others.

Note The example presented here uses various types of collections,

which might seem a poor choice because collections do not occur in
most domains—as collections. If we are seeking an object, we are likely
to be looking for something with a common domain name—a customer
or a product, for example. Collections are ubiquitous in every domain,
albeit with different names such as portfolio, bin, ledger, parking lot, air-
space, and so on. It doesn’t take very long before we recognize that any
object that exhibits the behaviors of adding elements, deleting elements,
ordering elements, returning subsets of elements, and iterating across its
member elements is either a collection or uses a collection as one of its
instance variables. As abstract as they are, collections soon become
some of the objects you use most and are most familiar with.
 Chapter 5 Vocabulary: Words to Think With 133

Always remember that the hierarchy is based on behavior and that classes
lower in the hierarchy are assumed to extend the behavior of those above them
in their branch of the hierarchy.

Abstract/Concrete
Abstract and concrete are labels for classes that do not have instances and those
that do, respectively. In creating a taxonomy, it’s convenient and sometimes
necessary to create a class solely for the purpose of representing (and in
software taxonomies, storing) behavior common to two or more other classes.
These classes are not intended to have instances. Concrete classes have
instances, so by definition, all objects are instances of concrete classes.
Abstract classes are always parent classes. An abstract class should not
appear below a concrete class in the hierarchy.
In languages lacking explicit taxonomic relationships among classes (in
C++, for example, where the classes are simply members of a library with no
explicit enforcement of the is-a-kind-of relationship, as is the case in Smalltalk),
an abstract class represents a “template” for a set of related classes. Conceptu-
ally it fills the same role: a convenient place for storing specification or imple-
mentation material that applies to a group of classes, which in turn will have
actual instances.

Inheritance
Inheritance is simultaneously a metaphor, a definition, and a mechanism for
implementing the definition. As a metaphor, inheritance suggests a family lin-
eage. The definition involves the establishment of an is-a-kind-of relationship
between classes. In some languages, such as Smalltalk, there is a “built-in”
mechanism that implements, independent of any application, inheritance auto-
matically when the is-a-kind-of relationship is declared. In other languages,
such as Java and C++, the programmer must assume more responsibility for
implementing the mechanics of inheritance.
There are actually three different inheritance metaphors, two suggested by
biology and one by economics. In biology, a child inherits the genes of both of
its parents and hence shares some traits and capabilities based on its parentage.
Biology also suggests the possibility of creating global structure that shows how
the is-a-kind-of relationship connects all living things. This structure is called a
taxonomy, and the one you are most likely to be familiar with was established
by Carolus Linnaeus. The third metaphor is suggested by our practice of allow-
ing relatives to inherit the resources of their parents.
All three metaphors have been used to explain inheritance. As discussed
in Chapter 4, object thinking uses the idea of a taxonomy tree based on an is-
a-kind-of relationship as long as characteristics and behaviors are used as the
134 Object Thinking

criteria for establishing the relationship. Object thinking explicitly rejects the
idea of basing a class hierarchy based on “DNA”—the internals of an object,
such as its methods or its instance variables. The definition of inheritance then
becomes, “A superordinate-subordinate relationship between classes in which
the subordinate (child) has the same behaviors (responsibilities) as the super-
ordinate (parent) plus at least one additional.”
The compiler or interpreter behind your implementation programming
language needs instructions on how to actually implement the inheritance rela-
tionship—how to search parent classes for methods or variables not physically
present in the child class. In some languages, this mechanism is automatic and
generally not accessed by the programmer; in others, it must be specified for
each program. Languages with implicit inheritance built in, such as Smalltalk,
have the advantage of an inheritance mechanism that has been optimized over
time and as a result of experience. Languages that expect the developer to
explicitly create the inheritance mechanism, such as Java and C++, will be as
effective as the developer is skilled. The tradeoff is greater control.
When classes have one and only one parent, they are said to participate in
a scheme of single inheritance. Classes may have a grandparent, great grand-
parent, and so on, but the entire lineage is based on the fact that a class has
only one parent on the next higher level of the taxonomy tree. Multiple inher-
itance implies that two or more parent classes exist for a given child class.
Figure 5-5 shows a fragment of a possible class hierarchy. Person, Student, and
Employee illustrate single inheritance, while WorkStudyStudent is an example
of multiple inheritance. The circles, aStudent, anEmployee, and aWorkStudy-
Student, represent instances of their respective classes. Person is an abstract
class and has no instances.
The italicized entries in the classes represent messages that can be sent to
objects of that class to obtain an object holding the information implied by the
message name. Send the name message to the aStudent or anEmployee object,
and you will get back aString, representing that object’s identification.
Inheritance can be visualized as follows. The salary message is sent to the
anEmployee object. The employee object does not physically contain a method
enabling it to respond, so the inheritance mechanism kicks in, and the object is
given a copy of the method defined and physically stored as part of the
Employee class structure. The anEmployee object then executes the method and
returns the expected string object.
In another example, aStudent is sent the name message. This time, the
inheritance mechanism does not find the requested method in the Student
class, so it looks in the parent class, Person, where the method is found.
Again, the aStudent object is able to execute the name method and return the
appropriate value.
 Chapter 5 Vocabulary: Words to Think With 135

Object

Person
name

single inheritance
an
Student Employee Employee
gpa salary
id id
aStudent
multiple inheritance

Work Study aWork-

Student Study-
department Student

Figure 5-5 Inheritance tree fragment, showing both single and multiple
F05MQ05

inheritance.

Inheritance is assumed to be intrinsically slow because it must execute

machine cycles to conduct its search for the appropriate method. The apparent
slowness is eliminated through optimization in all but the most severely time-
constrained applications. What happens if aWorkStudyStudent is sent the mes-
sage salary? A long search chain begins with a query to the WorkStudyStudent
class, the Student class, the Person class, and then the Object class, all with no
results. Having reached the top of the hierarchy, the search returns to Work-
StudyStudent and then to its second parent, Employee, where the method is
found and made accessible. Multiple inheritance increases the potential length
of the search chain and increases the time required to obtain the needed
method.
An even more problematic situation arises when aWorkStudyStudent is sent
the message ID. The needed method is not in the WorkStudyStudent class, so the
Student parent is queried and a method of the correct name is found, and aWork-
StudyStudent responds with a string representing a student identification number.
Now suppose that it’s 3:00 A.M., and aWorkStudyStudent is in the vault at the
Federal Reserve Bank. The requesting object is a security guard who wants to
make sure the student is authorized to be in the vault. When aWorkStudyStudent
returns her student ID, she is likely to be arrested. This situation can occur
because classes can have identical message signatures in their protocols. (See the
definition of polymorphism a little later on.)
136 Object Thinking

Multiple inheritance is unnecessary. Object thinkers consider it a bad idea.

It’s unnecessary because there are always design alternatives (such as delega-
tion, discussed in the next section) that remove the apparent need. It’s a bad
idea because it needlessly complicates the process of providing objects with
access to the methods that allow them to respond to messages. It’s a bad idea
because it introduces the potential for error and confusion when the wrong
mechanism with the correct name is used. It’s a bad idea because it necessitates
any object participating in a multiple-inheritance relationship to know
who might send a message (for example, the yourIdPlease message) and to
determine which of its inherited IDs (employee or student) it should return to
which client. Or the client must know about the internal structure of the
object it’s talking to and determine whether it should ask for employee.id or
student.id. In both cases, the communicating objects are tightly coupled—in
other words, they must violate the encapsulation barrier in order to determine
appropriate actions.
Object thinking will almost always provide alternative solutions to any
design problem that at first seems to mandate multiple inheritance.

Forward Thinking
Refactoring Stories
“Hi, Hector, we need to talk about these stories if you have some time.”
“Sure, glad to help. Should we stand over there by the whiteboard?”
“Yeah.” Hector and two pairs of programmers head for the white-
board. “Sally and Suroor are working on your story about updating
inventory, and June and I are working on the ‘menu displays available
products’ story,” Ron introduces the subject of the conversation. “During
the stand-up this morning, we started talking about the two stories and
thought we saw some overlap—for example, both the menu and the
inventoryReorderList need to know what products have been sold and if
any are left. So we went out to Vending Row and tried to visualize what
exactly goes on in both stories by looking at or imagining the objects
involved. We want to run our ideas past you and see if they are reason-
able and if they constitute some additional stories or modifications of
these two you gave us originally.”
“Sounds interesting. Shoot.”
“OK, I’ll start, and the others can jump in when they wish. Let me
draw on the board.” [Figure 5-6 shows the sketch Ron puts on the white-
board.] “Here we have the central object in these stories—the dispenser.

(continued)
 Chapter 5 Vocabulary: Words to Think With 137

Forward Thinking (continued)

Visualize the dispenser as a series of compartments, each one holding a

product. Over here we have the menu (a list of available items), and over
here we have the reorderList (a collection of line items, each containing the
product ID, the number sold, the number to reorder, and the average shelf
life information). Now, both the menu and the reorderList need to know
stuff about the products in the dispenser, particularly about the product just
entering or just leaving this first compartment in the dispenser.”

Figure 5-6 Hector’s rough sketch of the objects the group is talk-
F05MQ06

ing about, which he has made so that he has something to point to

while telling the stories.

“We all think there should be some kind of ‘stock dispensers’ story,”
interjected Sally. “We watched the guy from the vendor company loading
the candy machine, and we think the story would cover the reorder list
watching the dispensers—all of them and all of their slots. When a new
product is put in a slot, the reorder list asks that object for its productId
and adds a line item to itself containing that ID. It also asks the vending
machine for today’s date and time, which it gives to the averageShelfTime
object, which in turn will use it to calculate itself later. This story applies
to all the slots in the dispenser, even the first one.”

(continued)
138 Object Thinking

Forward Thinking (continued)

“But the first one gets contents in a second way,” continues Suroor.
“When a product is dispensed, the dispenser pushes the next product in
line into the first position. All the other products get advanced as well.
Now, the reorder list doesn’t care about this event—it already has the item
in its list from the time the vending machine was stocked. All it cares
about is the event occurring when a product is dispensed.
“The menu, on the other hand, does want to know which new prod-
uct has arrived in this spot. So it registers to be notified of any product-
Replaced event that occurs in the currentProduct slot. It then asks the
product coming into that slot for its name, which it adds to the appropriate
spot on itself for display.”
“When a product is dispensed,” adds June, “the inventoryList wants
to know as well. It actually needs to be notified of a readyToDispense
event so it can ask the product for its ID before it’s pushed off the cliff into
the retrieval bin. It then tells the line item for that product to add 1 to its
count of total sold. The averageShelfTime object also gets the system date
and time at this point so it can calculate a running average of the time that
that particular product ID stays in the vending machine.”
Ron takes the stage again. “When the inventoryList is asked to trans-
mit itself back to the home office, it tells all the line items to execute its
reorderAmountRule. Then all the information in the four fields is ready for
transmission: each product ID to be reordered, the amount sold since the
last time the reorderList was requested, the amount to be reordered, and
the average time on the shelf in that machine.”
“But that’s not part of our story—the two Sals are working on that one.”
Hector looks puzzled for a second. “Oh, you mean Sally and
Salvatore. Gotcha.”
“There is another special case,” Suroor says. “If a product is put in
the currentProduct slot and it has expired, the menu shouldn’t display its
name; it should display a ‘sold out’ message or something instead. We
don’t think that needs to be a separate story, but we wanted to check with
you to confirm our thinking about the problem.”
“Seems to me you have it nailed.” Hector ticks his fingers as he
relates, “It seems you are suggesting five stories: stock dispensers, update
menu, dispense product, update reorderList, and transmit reorderList.”
“Well,” Ron replies, “we already have a dispense product story; we
just need to make sure that the pair working on it know about the needs
of the Menu and the reorderList. And I think that the update reorderList is
just part of the need we should convey to the dispense product pair.”

(continued)
 Chapter 5 Vocabulary: Words to Think With 139

Forward Thinking (continued)

“OK, I’ll write the stock dispensers story and put it in the hopper for
the next iteration. We can just amend the narrative on the update inven-
tory story you already have to reflect our discussion. The same is true of
the update menu story. Those can stay in this iteration, then, and keep the
same priority—unless you think this will radically change your estimates?”
“No, if anything, it will make it easier, so we should finish in less time
than originally estimated when we didn’t really know what we were going
to do,” Ron says as the others nod in agreement.
“And,” continues Hector, “I will talk with the dispense product cod-
ers and make sure we have not changed their story so much we blow their
estimate. Good work, and thanks for the suggestions. I think this better
captures what happens and what we want in the final software.”

Delegation
Delegation is a way to extend or restrict the behavior of objects by composition
rather than by inheritance.
Consider an investment portfolio. A portfolio must be able to add and
delete investments, return a subset of investments (all the bonds, perhaps), iter-
ate across its members (asking each one its dividend date), and provide its cur-
rent value. All but the last behavior can be done by a collection object. So the
question arises, “Make portfolio a subclass of collection?” The answer, generally,
is no—give portfolio an instance variable, myInvestments, which contains a col-
lection object, instead, and delegate all the collectionlike behaviors of portfolio
to myInvestments. When you send a message like, “portfolio, what bonds do
you hold?” that question is delegated to myInvestments, which returns the bond
collection to portfolio, which returns it to you.
Delegation is one way to address the apparent need for multiple inheritance
in the case of a workStudyStudent. A workStudyStudent object could contain
within itself an instance of Employee and an instance of Student. It could then
define its own interface so that messages best handled by the anEmployee object
could be forwarded to the instance variable containing that object. The same
would apply to messages most appropriately handled by the aStudent object. Del-
egation eliminates the need for servers to be aware of their clients, and vice versa.
An object might want to advertise an ability to play different roles, and this
too can be addressed with delegation. The workStudyStudent might, instead of
creating its own protocol as a superset of Employee and Student, indicate that it
can assume either role upon request. Its protocol would then include two mes-
sages, asEmployee and asStudent, that would tell it to assume one of those two
140 Object Thinking

roles. All messages sent to workStudyStudent after that would be directly

received and responded to by the appropriate delegate object.
Delegation is used extensively in objectlike languages such as Visual Basic
and is a generally useful technique that was too often overlooked when inher-
itance was being stressed as fundamental to object implementation.

Polymorphism
When I send a message to an object, it has complete leeway to interpret that
message as it sees fit. The receiver of the message decides the appropriate
response and how to respond. As a sender of the message, none of that is my
concern. Because each object can interpret a message any way it wants (as long
as its protocol tells me what kind of object I can expect in return when sending
the message), more than one object can respond to an identical message.
Because every object has the right to define its own interface protocol,
it is not unusual for two or more objects to adopt the same message signatures.
A photograph and a document might both choose to implement a “print” mes-
sage. The details of printing will be quite different for the two objects, as will
the results. One message can yield many different responses, depending on
whom it was sent to. Thinking of each response as a form of response, we have
one message yield many forms (of response). The Greek for form is morph, and
poly used as a prefix means many—hence polymorphism (many forms). A bit of
a stretch, but that is where the term comes from.
Making the receiver of the message responsible for its interpretation gives
me a great deal of freedom in how I work with mixed objects. I can talk with
objects in ordinary vernacular. For example, I can ask them to print themselves
using a single message, “Please print yourself,” instead of remembering a different
message for each kind of printable object. A graphic object in the group will hear
the “please print” message and interpret it appropriately for a graphic, while a text
string will hear the same message and interpret it appropriately for a text string.
Polymorphism is a completely unremarkable phenomenon in the natural
world. We expect real-world objects to behave in this fashion and make jokes2
when situations arise in which objects respond to messages in unexpected
ways because they are polymorphic. Polymorphism is important in software
because it relieves me of the burden of coining an infinite number of message
variations so that I can satisfy the demand of a language compiler for making
every subroutine invocation unique.

2. There’s the joke about the large, leather-clad, menacing biker accompanied by his equally large and
fierce Doberman who enters an elevator car occupied by a middle-class couple. When the biker utters
the command, “Sit!” the couple as well as the dog immediately drop to the floor. The humor in this
story (if there is any) derives from the lack of polymorphism: the couple did not exercise their innate
ability to hear a message, determine whether it was for them, and react appropriately.
 Chapter 5 Vocabulary: Words to Think With 141

In the case of software objects, message signatures are supposed to capture

the essence of the message so that potential senders will know how to select the
appropriate message and object for their purposes. Good names are in limited
supply, and many objects in quite different areas of the class hierarchy will have
similar behavioral needs. Both students and employees have a need to identify
themselves, so it’s not unreasonable to allow them to have identical message
signatures to retrieve an identification object. It’s quite possible, in fact, for both
to have inherited a common message signature from a shared superclass. The
student object might respond with a string and the employee with an integer.
Even if both reply with a string, it’s likely that the string will have different values.

Encapsulation
The personal integrity of objects should not be violated. This is such an impor-
tant principle that the defining term, encapsulation, might better be considered
an essential term. I include it in the supporting terms category because of how
it is used—as an explanation and justification for other terms in the object
vocabulary.
Every object has a boundary separating public and private realms. This is
true whether the object is a software construct or a person. That boundary is
held to be impermeable: even though we know it can be penetrated, it should
not be. Encapsulation defines the “insides” (structural definition and enabling
mechanisms) of an object as private space, to or of which no one except the
object itself should have access or knowledge.
In most ways, encapsulation is a discipline more than a real barrier. Sel-
dom is the integrity of an object protected in any absolute sense, and this is
especially true of software objects, so it is up to the user of an object to respect
that object’s encapsulation. For example, even though it is possible to insert
wires into the brain of a human being and, by applying small amounts of elec-
tric current, make that human perform tricks, to do so would be considered a
gross violation of the person’s integrity. Users of software objects should dem-
onstrate the same respect.
Encapsulation implies more than respecting the public/private boundary.
Object users should not make assumptions about what is behind the barrier
either. This is true whether the assumption is about a piece of knowledge that
might be stored in the object or about the details of any message response
mechanisms that might be inside the object. Just because an object indicates
that it can provide a bit of information doesn’t necessarily mean that the object
has that information stored within itself—it might very well be obtained from
some other object without your knowledge.
142 Object Thinking

It is frequently the case that an object will be given responsibilities for

maintaining and providing bits of information that it does not possess. As a typ-
ical human object, for example, I have a responsibility to remember the names
and birth dates of my family members. I do a reasonably good job of remem-
bering the names because I have memorized them: they are somehow part of
my internal memory structure. I do not remember the dates. An external plan-
ner object, containing a page object, which lists my family members and impor-
tant dates associated with each one, is a very valuable collaborator. To an
external client, however, it would appear as if I contain both kinds of informa-
tion because I can respond to requests for both names and dates.

Component
A component is a large-grain object made up of several basic objects. A roof
truss is an example of a component. It’s made up of 2 × 6 pieces of lumber and
gang nails. It simplifies the process of constructing applications (in the case of
a truss, a roof or a house). A component has an interface independent of the
interfaces of its members.
Components, excepting GUI widgets, are frequently known as business
objects, and an example might be a checkbook. A checkBook object is com-
posed of a checkRegister and a collection of numberedChecks.
The distinction among an object, a component, an application, and a
system is somewhat arbitrary, however “clearly defined” advocates may
define each term. All can be called objects because all are packages of behav-
ior with an external interface and all, even the most basic object, might con-
tain other objects.

Framework
Framework is another term with multiple meanings. The four most common are
implementation (for example, a set of GUI widgets); foundational (a partial
solution to a problem encountered in multiple applications or domains), an
implemented pattern (for example, resource allocation); application, or vertical
market application (for example, banking); and architectural (for example,
client/server).

■ An implementation framework is a collection of classes that collec-

tively capture the behavior of a small, specialized domain. Examples
might include a graphics framework or a money framework. At this
level, the framework consists almost exclusively of a set of classes,
a small class hierarchy that can be added to a development
environment.
 Chapter 5 Vocabulary: Words to Think With 143

■ An abstract, or foundational, framework is both a collection of

classes and the scripts that guide their typical interaction. A foun-
dational framework offers an abstract solution to a particular prob-
lem that might be encountered in numerous applications across a
variety of domains. Examples of foundational patterns include
object routing and tracking, object allocation and scheduling,
object persistence, and many others. An object routing and tracking
framework, for example, would be useful in the construction of
applications as varied as package tracking software for a parcel
service and electronic network management software. Conference
center room scheduling, ticket sales, and airline seat management
applications could benefit from an object allocation and scheduling
framework.
■ Application frameworks are customizable solutions to common
domain needs. They are also known as vertical market frameworks.
Examples would include an inventory control framework, an
accounting framework, and a demand deposit framework. This kind
of framework is a completely functional software solution but one
that is easily edited (not reprogrammed) into a custom solution for a
specific client.
■ Architectural frameworks are the most general, and might better
be thought of as architectural patterns. Examples would include
client/server, model-view-controller, pipes and filters, blackboards,
presentation-abstraction-controller, and many others.

Pattern
Pattern is a term with one description and at least two meanings. The common
description is, “A named solution to a recurring problem-in-context that has
instructional value.” The original meaning of pattern came from Christopher
Alexander and reflected his focus on the problem space—a pattern was observ-
able in the world, it could be described, and the description could be used to
replicate the pattern when useful. The most common meaning of the term
comes from the book Design Patterns, by the Gang of Four (GoF)—Erich
Gamma, Richard Helms, Ralph Johnson, and John Vlissides—for whom a pattern
is a programming design solution to a recurring programming problem.
Most of the people using the term have been inspired by the work of
Christopher Alexander, an architect who proposed a pattern language of
design parameters that would allow the construction of anything from an “inde-
pendent region” to a montage of photos on the wall of a dwelling.
144 Object Thinking

Alexander’s goal was the discovery of the principles behind a “Timeless

Way of Building,” including structural, organizational, and aesthetic elements.
Alexander’s work tends to oscillate between the highly pragmatic and the
semimystical. It’s unsurprising that his work is subject to a wide range of inter-
pretation. In fact, the intensity of argument regarding what patterns really
should be is rivaled only by that of the original arguments about object
programming.
Richard Gabriel, James Coplein, et al. represent those who believe that the
semimystical aspects of Alexander represent the true essence of his message.
For them, a pattern—and even more important, a pattern language—is quite
different from what has been popularized as a pattern. The popular GoF view
presents patterns as elegant solutions to discrete design problems. Numerous
patterns of this sort might be incorporated into any given application.
Patterns for object thinkers are mental shortcuts or cues that direct think-
ing along known paths and facilitate the discovery of a problem solution. Of
course, this means that the patterns themselves must be consistent with object
thinking philosophy and principles, or their use will be counterproductive.
Patterns most useful to object thinkers should be derived from the prob-
lem domain, just as objects are. They should facilitate thinking about coordi-
nation and scripting of objects or about useful ways of assembling objects
into components or applications. They could be considered Alexandrian
patterns. Few of the patterns (about 6 of the 23) presented in the GoF book
satisfy this demand. Martin Fowler’s book on Analysis Patterns presents
examples derived from a domain and is much closer to Alexander’s intent
than the GoF book.
Patterns that reflect the solution space are useful to the object thinker but
in a very different way. If you have thought about and designed an object solu-
tion to a problem, you must still implement your solution using a programming
language. Not all languages directly support object ideas, and many languages
contradict object principles. If you must use such a nonoptimal language,
design patterns of the sort in the GoF book provide insights that can minimize
the distortion caused by the implementation language.
GoF patterns might better be called implementation patterns than
design (or thinking) patterns. Other types of implementation patterns would
include the coding standards that constitute one of the twelve XP practices.
In particular, programming idiom or style provides a powerful pattern for
implementation as well as for communication among developers. Apple’s
style guide for Macintosh GUIs, Kernigan and Ritchie’s style for the C pro-
gramming language, and Kent Beck’s Smalltalk Best Practice Patterns are all
examples of programmer idiom.
 Chapter 5 Vocabulary: Words to Think With 145

Implementation Terms
Those charged with the actual construction of software objects need vocabulary
to support their work—and support their object thinking—while dealing with
“the details.”

Method
Method is the name given to the block of code that is executed in response to
a specific message. Each message in an object’s protocol must have a corre-
sponding method.
As noted earlier, at this level it can be difficult to differentiate, visually
or syntactically, an object method from a traditional subroutine, function, or
procedure. Correct object thinking, however, will be reflected in the method
collective. As an aggregate whole, object methods will vary in significant ways
from a collection of routines or functions arrived at by applying traditional com-
puter thinking while programming. For instance, there will be fewer object
methods, and they will be simpler (on average) in their construction. Common
control structures, such as Case statements and explicit looping constructs, will
be absent (or if present, will be very few in number and will be used within an
object and not for object coordination or scripting).
Some of the methods used by an object might be considered private,
meaning that the object itself intends to use those methods and would prefer
that they not be invoked by other objects. Public and private are concepts that
apply to both methods and the messages that invoke those methods. An object
will frequently send messages to itself, resulting in the execution of a private
method, to obtain internally stored information or to obtain access to an object
created in another method. Some languages enforce a distinction between public
and private messages and their corresponding methods. Others do not, relying
instead on the integrity of programmers to respect the object’s design.3

Variable
A variable is a location or a container where an object might be located. Class vari-
ables are variables that are part of a class’s permanent structure. Instance variables
are variables that are part of an object’s permanent structure.

3. Languages such as C++ and Java also allow for protected messages/methods. This is a way to allow
outside objects to send such messages (invoke such methods) but restrict the number of outside
objects having that privilege. C++ also allows for friends, objects that have access to the private
messages/methods of other objects.
146 Object Thinking

Messages can be sent to a variable, in which case they are received and
responded to by the object residing in that variable location—the variable
becomes an alias for the contents (object in residence) of the variable.

Late/Dynamic Binding
This is another programming term that assumes a particular importance in the
world of objects. Type is a permanent4 label attached to an object—kind of like
a brand or a tattoo—that allows certain objects to be restricted to certain loca-
tions (variables) and allows a language compiler to enforce that restriction.
In some programming languages, the concept of type is minimized so that it
becomes important only when the program is actually executing. Those
languages are described as allowing dynamic, or late, binding. Not all program-
ming languages support dynamic binding, and whether or not they should is a
matter of much argument.
The value of dynamic binding lies in how it allows a developer to take
advantage of the natural polymorphism of objects in a direct and intuitive fash-
ion. It is possible, for instance, to create a variable, DisplayQueue, that will at
various times contain a text object, or a graphic object, or even a movie clip
object. I can send the display message to the DisplayQueue variable, and
whichever object is occupying the variable at that point in time will receive the
message, interpret it (polymorphism), and display itself. I do not worry about
the heterogeneous content of the variable.
A disadvantage arises from the possibility that I accidentally send an object
to DisplayQueue that does not know how to display itself, does not respond to
the display message. This can be prevented by asking any object seeking entry
to DisplayQueue whether it understands the display message. If it does, it’s
allowed in; if not, it’s refused entry.
The drawback of dynamic binding is the possibility that an inappropriate
object might come to occupy the variable, receive a message that it does not
understand, and cause the program to fail. In type-safe languages, erroneous
assignment errors of this kind will be detected by the compiler, flagged, and
corrected by the programmer before the program is allowed to execute. Unless,
of course, you cast types and get something like a Java “class cast exception”
error—the equivalent of Smalltalk’s “does not understand” error.
Solving this dilemma in a dynamically bound language is straightforward.
If the situation requires, I merely ask any object that seeks to occupy a variable
whether it understands a particular message or whether it’s an instance of a par-
ticular class, and I grant or deny residency based on its answer. This is extra
work, of course, and can have an impact on performance. Deciding when I

4. We are ignoring, for the moment, the possibility of casting types.

 Chapter 5 Vocabulary: Words to Think With 147

need to employ this kind of type checking is an important part of design and
of object thinking. The justification for allowing dynamic typing and inserting
explicit class membership checking when needed should be empirical.
Excepting the human interface, the probability of encountering a type error
should be quite low. If the consequences of error are minimal, the need to
insert extra code to check typing might not be needed. Simpler designs and
implementations should make implicit typing so obvious that programmer
error is minimized. The user interface will almost always have to enforce the
equivalent of type checking for every value entered, but once a value is “in
the system,” it should not need to be rechecked. Typing is sufficiently restric-
tive that programmers will employ features such as casting to escape the
restrictions. How many of which kind of errors, with what frequency, can be
prevented (or caused) by employing strict typing and allowing means for
avoiding type restrictions (casting)?
The merits of early and late binding are a source of considerable argu-
ment. We need not be concerned with that here. It must be noted, however,
that object thinking is better reflected in an environment that allows dynamic
binding. In the spirit that everything is an object, variables too are objects and
should have the responsibility of maintaining their integrity instead of giving
that responsibility to another object such as a compiler.

Auxiliary Concepts
The following terms do not define objects or concepts about objects. Instead,
they add nuances to, alter our understanding of, or enhance our perspectives
of those familiar terms.

Domain
The term domain refers to the arbitrarily bounded space we are simulating, in
whole or in part, with the objects we design and implement. We understand
objects based on how they reflect phenomena and concepts in the domain. A
domain might be a business enterprise or a type of business (for example,
banking or government). A domain might be nothing more than a focused com-
munity of objects collectively providing a particular set of services—for example,
the domain of graphics or the domain of money.
When we are constructing a class library or a set of components, our goal
should be to create a set that is capable of simulating the entire business enter-
prise or, preferably, the industry of which the enterprise is a member. It is
almost always a mistake to define the domain as coextensive with an applica-
tion program. A domain class library can be constructed incrementally as long
as every class added is a simulation of behaviors expected in the domain as a
148 Object Thinking

whole and not just the application in which a class is first encountered. Behaviors
can be added and moved via refactoring if the definition of classes is domain
driven. Yet this is the level at which too many object-oriented texts provide
illustrations and examples.
It is possible to define your domain as the computer—the implementa-
tion environment. This is, in theory, what object language designers do. If that
is your domain of interest, object thinking should apply as much to that
domain as it does to any other. Relatively few attempts have been made to
apply object thinking to these domains. Of course, it is precisely this area
where you might expect the greatest resistance to object ideas and the great-
est allegiance to machine thinking alternatives. Perhaps this is appropriate,
but it would be a lot of fun to seriously attempt the creation of an operating
system that reflects “pure” objects and nothing but objects. (There have been
attempts along this line, and Squeak, for example, has all operating system
services built into the class hierarchy so there is no need for an operating
system at all.)
This is probably a good point to insert a caveat about the material
presented in subsequent chapters of this book. Almost all of the examples and
discussions focus on application domains. Both an implementation domain
and an execution domain are assumed.
The implementation domain will be a programming language and its
accompanying class library. In most examples, we simply assume the existence
of objects such as strings, characters, and collections. Also assumed are typical
behaviors for those objects. Typical behavior is considered a superset of those
behaviors included in object programming languages such as Smalltalk, Visual
Basic, C#, Java, and C++. Sometimes, especially in dealing with collections, the
assumptions are biased by the capabilities of Smalltalk collections (and my
familiarity with that language), which are more extensive than in any other pro-
gramming language.
An execution domain consists of the virtual machine or compiler and the
operating system. Again, assumptions are made about services provided by these
entities. It’s also assumed that these entities are not generally object-oriented
environments. Because the operating system, for example, is not object-oriented,
it is frequently necessary to compromise object principles to some degree in
order to make it possible for objects to interact with nonobjects.
A special kind of execution domain would be a database management
system (DBMS). You could say that the very idea of a DBMS is antithetical to
object thinking,5 but such systems are an implementation necessity for most

5. Centralized hierarchical control and manipulation of passive data things: clearly, DBMS design, espe-
cially relational DBMS design, is not predicated on the kind of object philosophy and thinking dis-
cussed so far in this book.
 Chapter 5 Vocabulary: Words to Think With 149

organizations. Almost from the inception of object development, there has been
conflict between object applications and DBMS execution environments. The prob-
lem even has its own name: the impedance mismatch problem. (See Chapter 10 for
further discussion.)

Business Requirement
A business requirement is defined as any task, decision, procedure, or process
that supports a business objective, goal, or mission. A business requirement
might be satisfied by an individual object or by a group of cooperating
objects. Those objects can be human, mechanical, or software based. If the
business requirement can be satisfied by a single object, it becomes a responsi-
bility of that object. If a group of objects is required, the business requirement
is most likely to be expressed as a set of individual object responsibilities plus
a script that ensures the proper coordinated invocation of those responsibilities.
Most of the time we will use story, as used in XP, as a synonym for business
requirement.

Business Process Reengineering

Business process reengineering was first defined outside the world of objects
by Hammer and Champy.6 The inclusion of the term here is simply to provide
a bridge linking object thinking with business thinking. David Taylor illustrates
how object thinking can be applied to the business enterprise as a whole,
resulting in improved understanding and providing mechanisms for rethinking
the enterprise itself. There is another important link between the ideas of busi-
ness reengineering and thinking about objects. The better the object thinking,
the more likely the need to reengineer the business so that it matches the sim-
plicity, elegance, and flexibility of the new software. Object thinking almost
necessarily reveals better ways to do business, even when that thinking is
ostensibly directed toward software development.

Application
Application is the term used for a community of objects focused on accomplishing
a well-defined set of collective responsibilities. As used here, the term applica-
tion has almost the same meaning in object terms that it does in traditional
software development.

6. Hammer, Michael, and James Champy. Reengineering the Corporation, Revised Edition: A Manifesto
for Business Revolution. Harper Business, 2001.
150 Object Thinking

Note We have come to the end of our vocabulary list and have not
included one of the most mentioned terms in object literature—reuse.
There are several reasons for this. First, reuse is not a goal of object
thinking; composability is. Composable objects will be reused as a
matter of course, so reuse is but a byproduct of a more general goal.
Second, there are ways to obtain reuse that are not related to object
thinking—code libraries, for example—and the distraction is not really
helpful. Lastly, reuse was once touted as the premier benefit of object
orientation—a claim that proved to be highly overstated. Worse, per-
haps, was the claim that maximum reuse could best be obtained via
inheritance. Object thinking claims to lead to the discovery and craft-
ing of composable objects. The goal is to create a mindset that leads
to evolving flexible applications and systems that directly reflect and
support an application domain. Reuse will emerge, but it is not a
driving force.
Method, Process,
and Models
Caveats to keep in mind as you read this chapter:

■ All methods are someone else’s idea about what you should do when
you develop software. It may be useful, from time to time, to borrow
from those ideas and integrate them into your own style. It is essential,
however, to transcend any method, even your own idiosyncratic
method, and “just do it.”
■ Software development is like riding a surfboard—there is no process
that will assure a successful ride, nor is there any process that will
assure that you will interact propitiously with the other surfers sharing
the same wave. Published processes, like published methods, provide
observational data from which you can learn and thereby improve
your innate abilities—just as observation of master surfers enables you
to improve yourself.
■ No model has any value other than to assist in object thinking and to pro-
vide a means for interpersonal communication. If you can model your
objects and your scenarios in your head while engaged in writing code,
and if those mental models are consistent with object thinking, great! No
need to write them down. If you and your colleagues use a visual model
on a whiteboard as an aid in talking about scenarios and in clarifying
your collective thinking about those scenarios, and you erase the board
when you’re done meeting, also great! If your models are crudely drawn
and use only a subset of the syntax defined here (or a completely differ-
ent syntax that you and your colleagues collectively agree upon), still
great! Model when you must, what you must, and only what you must.
151
152 Object Thinking

For forty years, software development theorists focused on method, pro-

cess, and modeling as the key means for improving practice, based on the
belief that all software problems could be resolved if developers would give
up their idiosyncratic, imprecise, and careless ways. Formally defined methods
would incorporate rigorous process and modeling requirements. Following
the dictates of formal methods would eliminate all vestiges of ad hoc and sub-
jective “art.” Software development would be transformed, if not into a science
like mathematics and physics, then at least into a solid engineering discipline
solidly grounded on a science of computing. Models would have precisely
defined syntax (preferably based on a kind of predicate calculus), and all
semantics would derive from formal transformations of that syntax. Properly
constructed models would contain unambiguous truth, and one model (a data
flow diagram, perhaps) could be mechanically transformed into another
(source code, for example).
A process could be defined that would describe in precise detail each step
required to move from vague idea to functional solution. At each step in the
process, the developer would know exactly what to do, how to elicit required
information, and how to express that information in syntactically correct models.
Methods of this sort could be expressed as automated tools. It would be
possible to remove significant numbers of human developers from the
process—especially those pesky and annoying programmers. Computer-aided
software engineering (CASE) tools would allow a properly trained analyst to
construct precise models, which would then be transformed into error-free
code at the touch of a button (click of a mouse). It would, in fact, be possible
to build a repository, an ultimate CASE tool that would allow the automatic gen-
eration of computer programs and systems in response to a change in business
requirements—without the intervention of analysts or programmers.
The vision of the theorists was compelling—and highly salable to managers,
who rushed to adopt the latest advances, buy the latest tools, and pay for audits
that would document their compliance with the latest process standards.
Unfortunately for all, the dream was never realized in practice.
During the same period, practitioners—lacking such grand vision—focused
on discovering and sharing heuristics and practices grounded in experience
rather than theory. The art and craft followed by expert developers was transmit-
ted mostly via oral tradition, mentoring, and informal mimicry. Real improvements
in software development were realized—but non-self-consciously,1 in the form of
a shared culture and oral traditions.

1. Christopher Alexander differentiated between self-conscious and non-self-conscious processes in

architecture. The latter arise only in affluent societies with universities and give rise to architectural
theories whose merits are based on abstract meta-arguments rather than on any connection with
practice or actual manifestations. Alexander, Christopher. Notes on the Synthesis of Form. 1968.
 Chapter 6 Method, Process, and Models 153

After the fact, formal methodologists adopted many of these practitioner

innovations and recast them in terms of their pet theories (sometimes ignoring
the origins of their “insights”). This recasting allowed the methods of such
methodologists to give the appearance of supporting new innovations without
initiating any substantive change. In fact, innovations were all too often rede-
fined by the methodologists as being nothing more than some feature already
present in the method: “We have been doing objects since 1960; only we called
them ________.” This, and similar statements, were expressed by innumerable
mainstream developers and methodologists in response to innovations.
XP and the other agile methods represent the first attempt by practitioners
to systematize (not formalize) practice in such a way that it could be perceived
as a legitimate alternative to a formal method. So legitimate in fact, that the back-
lash from mainstream formal methodologists is intense. “XP is nothing more
than good software engineering.” “RUP (Rational Unified Process) and CMM
(Capability Maturity Model) are just as agile as anything coming out of the Agile
Alliance.” “XP works only in certain niches (which are not really important
anyway), while formal methods are required for most development.” “XP is
nothing more than a fad, a way for unemployed Smalltalk programmers to get
jobs again.”
The jab at Smalltalk programmers alludes to important relationships
among XP, methodology (in general), and object thinking. To fully see the con-
nection, it’s necessary to make one last historical foray—a brief recap of object-
oriented methods and models that have been advanced over the past 20 years.

Two Decades of Object Methodology

Scores of object development methods have been advanced since Alan Kay
coined the term object-oriented and applied it to software. The first methods
proposed clearly reflected the same presuppositions and philosophy that pro-
vided foundations for the Smalltalk programming language—specifically, an
orientation toward understanding objects in terms of their behavior. One early
method, by Adele Goldberg and Ken Rubin, originated within ParcPlace Systems
(a spinoff from Xerox PARC, which developed and marketed the first version of
Smalltalk). Object behavior analysis (OBA) was described in published papers
but was never truly marketed as a method. An automated support product for
OBA was developed but not, at that time, sold.2

2. Very late in the life of ParcPlace Systems (the spinoff that marketed the Smalltalk developed at PARC),
the OBA tool was sold—under the name MethodWorks. But by that time, it had no hope of capturing
any share of the development tools market. MethodWorks provided a set of forms, consistent with the
OBA models, containing textual documentation about the nature and relationships of objects.
154 Object Thinking

Behind the Quotes

Adele Goldberg
Adele Goldberg was a known figure in the Smalltalk world from the time
the language was conceived at Xerox PARC until it was eclipsed by Java.
Her concerns with education helped shape the evolution of Smalltalk,
and she contributed to its leaving PARC and becoming a commercial lan-
guage. When PARC spun off ParcPlace Systems, she became a board
member and later president of the company. With Ken Rubin, she
authored a behavioral object analysis method (object behavior analysis,
or OBA) and later a guide to object-oriented project management, enti-
tled Succeeding with Objects. While Goldberg was at ParcPlace Systems,
a version of Smalltalk with a much larger user base—from a company
named Digitalk—was acquired, and efforts were made to merge the lan-
guages and their customer bases. That effort failed, in part because Parc-
Place expended resources trying to sidetrack yet another competitor—a
company named OTI, allied with IBM—stopping the momentum of
Smalltalk (which was threatening to become the “next COBOL”) and
allowing Java (a pale imitation of Smalltalk at that time) to succeed.
Goldberg left ParcPlace and founded a company and product named
LearningWorks, which used the Smalltalk environment to teach comput-
ing and computer programming. For a time, the Open University in
England adopted the Smalltalk environment as a foundation for its
programming courses.

The year 1991 was a watershed year for object methods in terms of books
published. Booch, Coad and Yourdon, Jacobson, Rumbaugh, Schlaer and Mellor,
and Wirfs-Brock published books describing various methods. All of these
methods could be characterized as first generation. They were all produced
quasi-independently and exhibited significant syntactic variation.
A great deal of argument ensued. Which method was best? Much of that
argument coincided with arguments about implementation languages. It was
common for methods to be closely associated with languages and praised or
damned based on that association. For example, Booch’s method (which was
originally written to support Ada development) was preferred in the C++ com-
munity, while Wirfs-Brock was more popular with the Smalltalk crowd.
 Chapter 6 Method, Process, and Models 155

In addition to arguing with each other, methodologists took note of one

another’s work and recognized omissions in their own. Points of concurrence
were noted. Compliments were noted. Booch, for instance, specifically noted
the value of Class, Responsibility, Collaborator (CRC) cards for object discov-
ery and preliminary definition in the second edition of his method book.
Other efforts were made to extend the first-generation methods to support
specific types of problems: real-time systems and distributed systems, for
example.
At the same time, “The Market” demanded a single “right answer” as to
which method to adopt. Booch, Rumbaugh, and Jacobson, coming together
under the Rational corporate umbrella, were quick to consolidate their meth-
ods and models in order to present a unified method. Given the sobriquet
“The Three Amigos,” they advanced their cause. Their tool sets and methods
became the foundation for Unified Modeling Language (UML) and the Rational
Unified Process (RUP)—a tool and a method that dominate “official” 3
practice today.
A review of all of the methods advanced in the nineties is not possible
in this space. It is possible to group the plethora of methods into three general
categories (data-driven, software engineering, and behavioral) and look at
exemplars of each category in order to outline the main points of divergence.
Data-driven approaches are exemplified by the work of Schlaer and Mellor,
the Object Modeling Technique (OMT) of Rumbaugh et al., and the joint
work of Coad4 and Yourdon. The central focus of this category of methods is
data and the distribution of that data, via the classical rules of normalization,
across a set of objects. Objects in this instance are equivalent to entities in
classical data models. Schlaer and Mellor, for example, describe the process for
discovering objects in a manner that is indistinguishable from data modeling
techniques.
OMT syntax and models allow greater variation and flexibility in rela-
tionships among objects than do the models of Schlaer and Mellor, capturing
interaction among objects in non-data-specific ways, but they still define
objects in terms of data attributes. Coad and Yourdon’s object-oriented analy-
sis (OOA) method essentially advocates the creation of an entity diagram,
placing functions with the distributed data attributes and adding message
passing to the data model.

3. A huge discrepancy remains between what developers officially use and what they actually use, in
terms of both method and model. Licenses sold still does not equate to actual use by developers.
4. Coad’s later and independent work was far more focused on behavior and patterns of behavior among
groups of objects.
156 Object Thinking

Data-driven methods do not bring about the object paradigm shift.

Instead, they carry forward a legacy abstraction, data, and use it as the basis for
decomposing the world. Followers of this type of method “think like data” or
“think like a (relational) database.” This obviously has the appeal of familiarity
and consistency with legacy systems and provides the appearance of a quick
path to objects. It’s also quite consistent with the definition of an object as a
package of data and methods.
It isn’t consistent with decomposition of the world in a natural way
because the world isn’t composed of computationally efficient data structures.
There are no “natural joints” in the domain that map to normalized entities. This
disjunction is evident in the world of data modeling itself: just consider the
differences between a conceptual data model (models data as understood
in the domain) and a logical data model (models data as normalized for
implementation).
Data-driven methods tend to create objects with more frequent and tighter
coupling than do other methods. Class hierarchies are therefore more brittle,
and changes in class definitions tend to have greater impact on other class def-
initions than desirable. (See Figure 6-1 for an example.) Distributing data across
a set of objects in accordance with normalization rules and by following the dic-
tum that a class must be created if a class (entity, really) has at least one
attribute different from any existing class needlessly multiplies the number of
classes required to model a domain. (See Figure 6-2.)

Owner Legal Event Vehicle

name dateTime VIN

phoneNo fee Year

calculateFee
acceptFee Passenger
Fee was an attribute of Title and Registration -
moved to Legal Event superclass.
Truck
calculateFee and acceptFee methods placed
with class containing data. Both are coupled to Title Registration
entire Vehicle hierarchy. Code must therefore be Motorcycle
implemented with case statement and modified
every time the Legislature decides to implement Trailer
a new tax rate or vehicle type to be taxed.

Figure 6-1 In this Department of Motor Vehicles example (patterned

F06MQ01

after Coad and Yourdon, 1991), data modeling rules lead the designer to
place the fee attribute and therefore the calculate fee method in the Legal-
Document class. This mandates the need for a case statement to handle
fee calculations and couples LegalDocument to the entire Vehicle class
hierarchy.
 Chapter 6 Method, Process, and Models 157

Customer

custId Is-a-kind-of
fName
lName
MI
CreditCustomer
The creditLimit attribute “does not apply”
to all customers. By the rules of normalization cCustld
it must be moved to another entity - a subtype creditLimit
of customer. Entities proliferate needlessly.

Figure 6-2 In this example, the addition of a single new attribute

F06MQ02

requires the creation of a new class, allowing some customers to buy on

credit and therefore have a creditLimit attribute, for example.

Booch and Jacobson exemplify softwar e engineering methods.

They openly acknowledge an intellectual debt to classical structured devel-
opment and claim to be an appropriate response to the demand by manage-
ment and government for a documented (formal) process for software
development. They are characterized by multiple models that, when (and if)
integrated, present a comprehensive specification of a desired software
artifact. Supporting code generation is a frequently stated motive for these
methods.
Most software engineering methods claim to be (partly) behavior driven.
Jacobson’s use cases are essentially identical, in fact, to the scenarios of OBA, a
prototypical behavioral method. Because use cases are focused on discovering
requirements to be used to define an artifact—the software—they tend to fall
into thinking about an object as encapsulated data and procedures: they reflect
thinking like a computer.
Nothing in these methods compels a design to reflect computer think-
ing. However, replication of modeling syntax from structured development
and the vocabulary employed in discussion (for example, attributes and
operations) reflect the artifact-centric point of view borrowed from struc-
tural development. Developer focus remains on the artifact and its design to
meet specification, rather than on understanding and modeling the problem
domain.5

5. Only nominal attention is paid to the process of developing a domain class hierarchy as a tangible
product in most software engineering methods.
158 Object Thinking

Software engineering methods (with the exception of Booch’s 1991

book) fail to stress the paradigmatic differences6 between objects and tradi-
tional modules and data structures. Even use cases, as developed by Jacobson,
are less about responsibility discovery and assignment than they are about
requirements gathering.
Behavioral methods, despite being first and most closely allied with the
development of object thinking, have always received less attention than data
and engineering methods. Proponents of other methods are quick to assert that
this lack of popularity reflects fundamental flaws or lack of utility in behavioral
approaches. They are wrong: other factors account for the failure of behavior-
alism to capture market share.
First, behavioralism was not promoted, as were competing methods.
Important advocates of behavioral ideas—Goldberg and Rubin at PARC—and
the originators of the most widely used behavioral modeling technique—Beck
and Cunningham and CRC cards—never published any kind of method book.
At a time when management almost demanded an automated tool to accom-
pany and support any new method (this period of time was also known as the
era of CASE tools), the following occurred:

■ ParcPlace Systems chose not to market the tool (OBA) it had created,
although it did provide the tool to those taking its object analysis and
design seminars.
■ Beck and Cunningham actively lobbied against the creation of any
kind of automated CRC tool and, today, argue against automation of
story cards in XP. Their objective was quite sound, philosophically,
but disastrous in terms of marketing.
■ Knowledge Systems Corporation developed a tool named coDesign,
based on Smalltalk, that was highly regarded by those that saw it
demonstrated. It never became a commercial product because of
financial and political concerns within KSC.
■ Corporate information technology organizations were actually mov-
ing to adopt Smalltalk at this time but refused to give up their rela-
tional databases, so tool vendors concentrated on object-relational
mapping tools instead of behavioral modeling tools.

6. Booch is a notable exception to this rule. He clearly recognizes that a fundamental shift in thinking is
required for object development. He also clearly recognizes the need to decompose the world based
on the domain expert’s point of view. He parallels Parnas’s views on domain-centric (design-centric)
decomposition. His method, however, does not develop or directly support this aspect of object
development. Instead, his models and syntax are directly focused on what is needed to construct the
software artifact.
 Chapter 6 Method, Process, and Models 159

The first, and for a long time only, book promoting the behavioral
approach to objects was that of Wirfs-Brock, Weiner, and Wilkerson. Nancy
Wilkinson published a small volume outlining classical CRC cards a few years
later—long after the opposition had all but won the “method war.”
Second, and more important, behavioral approaches were the least devel-
oped in terms of expressiveness of models and making the transition from analysis
to design and implementation in a traceable fashion. It would not be misleading to
characterize behavioral methods, circa 1991, as “CRC cards—then Smalltalk.”
Wilkinson, in fact, suggests that the primary value of CRC cards and
behavioralism is to provide an informal and rapid way to obtain input for more
formal software engineering methods. Booch, OMT, and OOSE (Jacobson’s
Object Oriented Software Engineering) were the formal software engineering
methods she seemed to have had in mind.
Third, behavioral approaches are the most alien to established software
developers—even today. Thinking like an object is very different from traditional
conceptualizations of the software development process. It requires, at least at
the beginning, constant diligence to avoid falling into old mental habits. It is hard.
Most people are unwilling to engage in this kind of hard cognitive work without
a compelling argument as to why it is worth their while. That argument was never
made—except in a kind of oral tradition shared by a very small community. This
book is an attempt to capture several aspects of that oral tradition. As with XP, the
primary justification for object thinking derives from a better fit between informa-
tion technology and business and significant reductions in complexity.
In fact, behavioralism had the same reputation as XP does today—as being
anti-method. There is a small element of truth in this assertion. Both OO and XP
proponents oppose the use of methods as traditionally defined—especially
comprehensive, highly formalized, and labor-intensive methods that prevent
developers from immediate engagement with code—with programming.
This does not mean that object thinking lacks anything resembling a
method. (The same is true of XP.) It merely means that the rigor and the sys-
tematization of work that accompanies object thinking is quite different from
what most developers have come to associate with the term method.
To understand the relationship between method and XP, it’s necessary to take
a few moments and reflect on how method provides value to software developers.

Purpose and Use of Method

Methods come in many guises. The simplest is nothing more than the adage, “If
at first you don’t succeed, try and try again.” Hacking, trial and error, and rapid
prototyping are examples of methods based on this precept. Iterative develop-
ment approaches—based on exploration, trial and error, and incremental
160 Object Thinking

expansion from small working systems—have been used and advocated almost
since the inception of commercial computing but have never been accepted as
“official” methods.
Other methods are more prescriptive, instructing the developer as to what
to do and when to do it. The simplest form of a prescriptive method would be
a checklist, such as that used by pilots for each phase of flight. Such checklists
consist, essentially, of a set of questions of the form, “Did you remember to do
this action?” Actions that need to be done in a particular order simply have their
reminder questions occur in the necessary sequence.
Most software engineering methods are highly prescriptive—with precise
instructions as to what and how to develop software at each meticulously
defined step in a carefully arranged process. They are based on a value system
that incorporates a basic mistrust of the human worker. Such methods can
become highly complex and comprehensive as the checklist expands to include
documentation that must be produced to confirm that each action did in fact
take place, as well as syntax that specifies the exact form of the documentation
so that it is known that the action was done correctly. (I have seen one such
method that filled almost 20 volumes—defining each step, describing each doc-
ument or artifact produced, and including algorithms for determining the cor-
rectness of all artifacts produced.)
Between hacking and prescriptive overkill is the realm of formal, informal,
and aformal methods.
Formal methods tend to be prescriptive. They are characterized by a large
number of models each of which requires a reasonably complex syntax, they
are filled with rules and techniques that assure proper use of the method, and
they are fairly strict in terms of activity sequencing. Most offer a promise of
removing human developers, to some degree, from the development process—
via code generation, for example.
Informal methods require fewer models with simpler syntax, demand
fewer activities, and offer heuristics instead of rules and techniques. Most pick
a single activity—programming, for example—and elevate it to primary status
while suggesting that all other activities are valuable only insofar as they
support the primary status. XP shares these characteristics and would be called
an informal method.
An aformal method7 would reject the idea that any task, model, syntax,
rule, technique, or heuristic has any intrinsic value. Such things are valuable
only to the extent that they support or assist the innate capabilities of the
human developers engaged in a particular task at a particular moment in time.

7. Aformal method is almost, but not quite, an oxymoron. Some familiar aspects of method remain even
in the most aformal approaches to software development.
 Chapter 6 Method, Process, and Models 161

Both formal and informal methods are seen as a kind of exoskeleton that might
offer some protection, but at the cost of severely limiting the being “enjoying”
that protection. Improving the innate character and capabilities of the human
developer is the alternative to defining an external method, for an aformalist.
Aformalism might also be considered “master practice.” Craftsmen have inter-
nalized those aspects of informal method and practice most germane to their art
and have transcended them.
An analogy to chess playing might illuminate the relationships among
method categories. A beginning chess player follows rules and defined proce-
dures (they are formalists), while a journeyman (informalist) relies on patterns
and heuristics. A grandmaster has internalized and transcended the informal to
become an aformal player.
Methods themselves are less important than the culture shared by those
that embrace a method. Software engineers, for example, represent a culture
enamored of formal methods (even if they cannot use them). (There are excep-
tions to this rule: Dave Thomas, for example, sees engineering as creative and
aformal problem solving independent of any formal method or process. It is
hard, however, to find that attitude reflected in software engineering texts.)
Those methods are considered necessary and desirable because the method
expresses the set of presuppositions, values, and worldview shared by the
members of the culture. Formalists like formal methods—an obvious truism.
Managers; academics; software engineers; computer scientists; and propo-
nents of UML, RUP, and CMM all tend to be formalists. Practitioners, as Robert
Glass has shown, generally are informalists. XP and object thinkers aspire to be
aformalists.
Methodological conflicts, therefore, are really cultural conflicts. When an
advocate of RUP asserts that “RUP is agile,” he or she is making an irrelevant
statement. Of course, RUP can be agile—it is merely a tool, after all, and the
practitioner has the choice of how, when, and why to use that tool. But some-
one from the culture that values RUP cannot be agile. (OK, technically, will not
be agile. OK, OK, is very unlikely to be agile.) The same mostly nonconscious
worldview that leads to adoption of RUP and UML will prevent the use of that
method in any but a formal way unless a constant, conscious, and deliberate
effort is made to adopt the agile worldview and cultural values at every step of
development.
Formalists will advocate methods that are focused on the production of
computer artifacts (software programs). Such methods will liberally employ
terms such as software engineering and will stress correctness and both syntac-
tic and semantic integrity. It is held that such methods are capable of producing
a “correct” solution to an unambiguous requirement, prior to the expensive
process of committing it to code. Code generation is then a happy byproduct of
162 Object Thinking

formal methods. The major drawback, of course, is that the effort required to
produce the formally correct models is essentially the same as (or more than)
that required to code and test. (Code, ironically, is far more amenable to formal
testing than any abstract model. As Ken Auer says, “Bubbles and arrows never
crash; they also never run!”) This makes for a steep learning and overhead curve
for those using this kind of method. They will also suffer from some degree of
mismatch between the real world and the formal world they attempt to
describe. The real world is far fuzzier, flexible, and illogical than any formal
method can accommodate.
Informalists are more likely to focus on methods that support negotiated
understanding—for example, the informal communication-oriented methods such
as joint application development (JAD) and CRC. To the extent that they adopt
modeling techniques, informalists see the models as a means of communication—
not an artifact with intrinsic value. It is clearly recognized that such models are
bound in terms of time and are meaningful for the group involved in their
creation. That is why the development group must include users, managers,
analysts, programmers, documenters, maintainers, and so on. They will
understand the models only if they participate in their construction. Reinhard
Keil-Slawik 8 shows how such models (including code) serve as a kind of
external memory for the group that created them and works with them. The
primary problem with these methods is the lack of technique for translating
the human understanding of a solution into a “computer understanding.” The
computer, after all, is a formal machine and can operate only on formally
defined products. This explains the tendency of informal methods (XP, rapid
prototyping, open source, and so forth) to focus on programming—the actual
interface between formal machine and informal developer.
Aformalists appear to reject method entirely. Observation, however, reveals
patterned behavior and tasks performed in a sequential manner. Models are
constructed (on whiteboards, Post-It notes, and 4 × 6 cards more often than with
the use of an automated tool), giving those models a relatively short half-life.
Aformalists do not reject method so much as they reject the idea that methods and
models have any intrinsic value apart from those using them in a specific context.
In one of the Alien movies, Sigourney Weaver’s character fought the alien
monster using a mechanical exoskeleton. This exoskeleton greatly increased
her human strength by amplifying her human movements with electromechan-
ical means. Formalists tend to see method as a similar kind of exoskeleton, one
to be wrapped around a human developer in order to amplify his or her skills.

8. Floyd, Christiane, Heinz Zullighoven, Reinhard Budde, and Reinhard Keil-Slawik. Software Develop-
ment as Reality Construction. New York: Springer-Verlag, 1992.
 Chapter 6 Method, Process, and Models 163

Aformalists viewing this scene would be mildly impressed by the mag-

nification potential of the exoskeleton but would be completely appalled
at the cost required to gain that advantage. The human becomes a mere part
in the machine and is severely restricted to a range of actions determined by
the exoskeleton and its designers. They would note that continued use of
such mechanical strength enhancement devices would perpetuate the physi-
cal weakness of the human and, likely, increase those weaknesses (exagger-
ating the importance of the exoskeleton) over time. In addition to the
limitations imposed by the designers of an exoskeleton, use of one adds new
complexities—how to get in and out efficiently in response to perceived
need—and vulnerabilities—what to do when your exoskeleton is in the next
room when the alien appears?
Object thinkers and extreme programmers reject the method-as-exoskeleton
approach in favor of a method–as–weight room idea. A human uses mechanical
weights and machines in a gym to increase his innate capabilities—to make his
own muscles stronger and more reliable. Using a more cerebral metaphor (and
one therefore more appropriate for object thinking), Kent Beck suggests using
method and even XP practices as if they were etudes (musical exercises
designed to help the musician internalize certain skills and techniques). Etudes
are used for practice, to increase and discipline the innate capabilities of the
musician so that she can then go on stage and perform music.9
Object thinkers and extreme or agile developers find value in method only
to the extent that it helps them become better practitioners. The method itself
should wither away as developer skills increase until it becomes the merest ves-
tige required to act as a reminder or trigger for the human developer to apply
his or her enhanced skills when and as appropriate. Kent Beck suggests that
there are three XP maturity levels: out of the box, adaptation, and transcen-
dence. These levels correspond nicely with the formal, informal, and aformal
typology introduced earlier in this chapter. (Except XP would never have a
formal level—substitute rote performance for formal in the XP context.)
This is not a new idea. The south Asian philosophy of karma yoga is
grounded in the idea that right actions allow you to minimize the consequences
of your actions sufficiently that you can maximize your chances of enlighten-
ment. Aristotle suggested, “Wear the mask of a good man long enough and you
become that good man.” David Parnas suggested some very good reasons for
faking a rational design process even when you can’t actually follow such a
thing. Christopher Alexander stressed repeatedly that a pattern language is but

9. It is true that some etudes are themselves sufficiently beautiful that they get played in concert, just like
any other kind of music composition. This does not detract from the intent behind their creation.
164 Object Thinking

a gate that must be passed through and left behind before you can actually
practice the “Timeless Way of Building.”
For object thinkers, method is never an end in and of itself. There is no
intrinsic value in either method or process. Both are useful only to the extent
that they provide practice in the use of, and a means of enhancement for,
innate human capabilities. Within the context of this worldview, it is possible
to ask whether some methods, some tools, some models are more suited to
promoting object thinking than others. And the answer is yes. To determine
which methods, tools, and models are the most efficacious, the following criteria
can be used:

■ The method, tool, model, or process cannot substitute its own goals
in place of those articulated in object (or XP) philosophy.
■ Each task advocated by the method and each model created using
the method must contribute to the realization of basic goals.
■ The models, vocabulary, and syntax associated with the method and
its models must value expressiveness over correctness. Models must
evoke knowledge in the head of the developer instead of making
pretensions to unambiguous representation of that knowledge.
■ The methods and models must
❑ Provide support for “natural” decomposition and abstraction of
problem (enterprise) domains. This requirement refers back to
the simulation notion behind SIMULA and the object paradigm.
❑ Recognize a need for two complementary processes: one
focused on domain modeling and the second on application
assembly.
❑ Include heuristics for discovery and evaluation. Heuristics
should be grounded in appropriate metaphors to facilitate
learning.
❑ Include heuristics or metrics that allow you to measure
progress and the “goodness.”

A Syncretic Approach
The preceding discussion emphasizes the differences between formal and afor-
mal approaches to software development. It might lead one to suspect that the
gulf between the two philosophies is unbridgeable—especially when consider-
ing method.
 Chapter 6 Method, Process, and Models 165

Even hard-core object purists recognize that there are many things of
value even in the most formal of methods. Blending methods and approaches
presents a significant challenge.10 Incorporating valuable ideas from formal
methods in such a way that the worldview and values of object thinking are
preserved is equally hard.
Nevertheless, it is desirable to find some kind of common ground. Two
prerequisites are required. First, the term method will be abandoned in favor of
approach. This is purely a political move—to eliminate distractions caused by
the “M-word” itself. Second, we will borrow a concept—syncretism—from
anthropology and religious studies as a metaphorical springboard for develop-
ing a syncretic approach.
Syncretism refers to a particular type of blending of traditions and cul-
tures. For example, a visitor to a Catholic church in the Caribbean or South
America will almost certainly find representations (icons, statues, relics, feast
days, and so on) of Catholic saints that are indistinguishable, on the surface,
from “pagan” deities. These vernacular adoptions do not change the underlying
principles of Catholicism; they merely make those principles more accessible
(as in the use of vernacular in the mass) by interpolating a mediating form. A
syncretic approach to software development would be characterized by a
smooth transition from behavioral decomposition and analysis into representa-
tions sufficiently formal that they can be implemented as computer software.
(In XP, this transition takes place in small steps, one story at a time.) Along the
way, the user would see many elements borrowed from other sources, ele-
ments that might even be redefined in various subtle ways to maintain consis-
tency of expression. It will be quite possible to maintain object thinking and XP
practice while drafting UML models that are syntactically, but not semantically,
identical to UML models drafted by a traditional software engineer.
Using the label syncretic approach is deliberate. Although many will take
what is advocated in this book as YAM (yet another method), the intent is to
create an approach to thinking about objects that nevertheless allows adoption
of techniques, insights, or models that may have originated in traditional com-
puter science and software engineering. Forms may be adopted from formalist
methods, but never underlying principles or philosophies. Succeeding chapters
will develop the syncretic approach more fully, but the remainder of this chap-
ter will provide a summary of the process and a quick definition of the models
(all of which are adapted from existing methods) that will be employed.

10. UML and RUP, it has been reported, took a lot of compromise and negotiation before they were
hammered out, and the three methods synthesized there were basically similar. The elements of
behavioralism in Booch were essentially discarded to make the synthesis possible. It proved much
harder to combine behavioralist and software engineering approaches, although a product named
Fusion, from Hewlett-Packard, made a valiant attempt.
166 Object Thinking

Paradigms, metaphors, and concept definitions provide the foundation for

understanding object orientation. They do not, however, tell us much about
how to do object development. “How-to” requires the following:

■ The enumeration of specific actions that can be taken to reach

certain goals.
■ Agreement as to a vocabulary (both verbal and graphic) and a grammar
(language) for assuring mutual communication and understanding.
Style, idiom, and standards are important aspects of this common
language.
■ Criteria by which we can evaluate ourselves, our progress, and our
products.

Object thinking imposes four additional requirements that must be

addressed by a syncretic approach to development. Implicit in these four
requirements (listed next) is the demand that two separate but complementary
processes (as discussed in Chapter 4) be respected—object discovery and spec-
ification (Lego brick construction) and application assembly and scripting.

■ Decomposition based on discovering the “natural joints” in the

domain. What goes on inside a computer, the implementation con-
text of the software to be created, is not part of the domain.
■ Responsibility assignment based on expected behavior in the domain
as observed or inferred in the domain. Preserving existing channels
and means of communication is essential. Eric Evans11 has just pub-
lished an important book detailing why this emphasis on domain-
driven modeling is so essential. Thinking like a computer, even when
you are creating tests and writing code to realize a single XP user
story, should not play a role in defining objects (decomposition of the
story) or defining and assigning responsibilities to those objects.
■ Aggregation of objects into communities capable of interaction and
collective solution of a task or tasks. Not only must you identify what
must be done by a group of objects acting in concert, you must also
identify the constraints under which they must work and any qualita-
tive characteristics that will be used to evaluate their work. Interaction
of both human and automated objects must be taken into account.
■ A reasonably direct metaphor-preserving means of adding design
and implementation detail to objects based on their responsibilities.

11.Evans, Eric. Domain Driven Design—Tackling Complexity in the Heart of Software. Boston: Addison-
Wesley, 2004.
 Chapter 6 Method, Process, and Models 167

The most tangible elements of a syncretic approach satisfying these

requirements will be the set of models suggested for use by developers. This
chapter will conclude with a brief introduction of a set of models and their syntax.
Subsequent chapters will explore the actual use of those models.

Forward Thinking
Review
One Friday afternoon the team is getting ready to leave, when Sally looks
around the room as if seeing it for the first time. “You know,” she says
almost to herself, “this place is a pretty good visual metaphor.”
“How so,” asked Ron, who was standing nearby.
“Well, there are the obvious things, like all the charts on the
wall, the bulletin board with story cards; they are visual metaphors for
communication.”
“A lot of projects have walls full of charts and diagrams.”
“Yes, but ours tell the truth. People believe them because they show
both good and bad—just the facts ma’am,” Sally said with a rather bad imi-
tation of Jack Webb as Sergeant Friday. “Our documents are not fancy
or polished, but they are useful—pragmatic aids to communication, not
valuable artifacts in themselves.”
“They’re playful, too,” suggested Ely as he joined the conversation,
“like those object cubes that Dan made by taping three-by-five cards
together. Remember the way that Suroor and I were juggling them as we
talked through that problem last week?”
“And they can be stacked up—a metaphor for how you can build
things with objects. As useful as they were, the old CRC cards could not do
that easily. Building a house of cards is a lot harder than a house of blocks.”
“A cube is a pretty simple shape—and XP prizes simplicity.”
“It would take a videotape to show it, but the way our walls came to
be covered in paper would contribute to the metaphor as well. When we
started, the walls were bare; it was only as our knowledge of what we
were doing increased that we had a need to externalize our group mem-
ory with the diagrams and object cubes.”
Most of the team had joined the discussion by now, and Sally was
looking around the group. “There is something else, maybe the most
important thing of all—you. If you look at the people in this room, you
see something very different than you would in any other software shop.

(continued)
168 Object Thinking

Forward Thinking (continued)

It’s in everyone’s body language and expressions and interactions. Hard to

describe but very obvious. Everyone here has confidence—we all have
internalized a lot of knowledge, not just about our project, but about
object thinking and XP practices. It’s a part of us as individuals.”
“I know what you mean, and it’s not just because we learned a few
new programming tricks, or gained some experience, or even because this
project is wildly successful. Somehow we have learned a new way to
think about problems, a new vocabulary to think with, and a new way to
perform—to work.”
“And joy and serenity along with the confidence.”
“Hah, sounds like we are all enlightened somehow.”
“I think we are, at least in one important aspect. We have learned
a lot about thinking, and we have applied it and we have somehow
transcended it—as with Christopher Alexander’s pattern language, we
have passed through the gate and are living object thinking.”
“What should we do with this enlightenment?”
“When the Zen monk was asked how he exhibited his enlightened
state, he replied: when I am hungry I eat; when I am sleepy I sleep.”
“And how does that apply to us?”
“Well, when Monday comes we will all come in and write software.”

Models
Models are valuable only to the extent that they facilitate communication
among human beings. (A model that can be “understood” by a computer is a
programming language.) This implies that the effort required to construct the
model must be less than the communication value that arises from its use. The
dramatic failure of CASE can be traced almost entirely to the fact that the effort
required to learn and use the tools far exceeded the ability of resultant models
to facilitate communication.
Unfortunately, a learning curve and some degree of overhead will always
be associated with the use of any set of models and any automated tool
 Chapter 6 Method, Process, and Models 169

supporting the construction of those models. It’s the balance between benefit
and cost that’s critical in determining whether models and tools will be useful.12
In that spirit, none of the following are essential in the sense that you cannot do
object thinking without them. None of the models have fixed syntax or content—
what is presented is suggestive. All of the models can and should be simplified
as the developer internalizes object thinking skills—until, as with the Cheshire
cat, only the model’s “smile” remains to evoke the knowledge in the head of the
developer and of the development team.

Semantic Net
A semantic net consists of little more than nodes, arcs, and text labels (see
Figure 6-3), with nodes representing things and arcs representing relation-
ships among those things. Nodes are labeled with a noun or noun phrase and
relationship arcs with a verb or verb phrase.

verb phrase
Noun
Phrase

Noun
verb phrase
Phrase

verb verb phrase

Noun
phrase
Phrase
Noun
Phrase
Noun verb
Phrase phrase

Figure 6-3 A semantic net, a brainstorming tool for discovery of potential

F06MQ03

objects in a domain.

A semantic net can be useful to jump-start a discussion of objects in the

same way that underlining nouns and verb phrases in a written problem
description. In both cases you are attempting to quickly identify potential
objects and typical—in the domain—interactions among those objects. The pri-
mary utility of a semantic net is to serve as a brainstorming tool that can be used
by a group of people simultaneously. (A good-size whiteboard is very useful.)

12. Booch clearly recognized that developers had an upper limit on the amount of time they were willing
to expend learning a method or syntax. He advocated the use of a subset of his full modeling set,
called whimsically “Booch Lite,” as the kernel that was most useful most of the time.
170 Object Thinking

The model produced need not persist beyond the point when the objects,
responsibilities, and constraints identified have been transferred to other
models (such as the object cube). The LIMT sidebar illustrates a typical use of
a semantic net.

LIMT—Low-Income Mortgage Trust

The Legislature has created a $2 billion dollar fund—the LIMT—to help
young couples acquire a home they ordinarily could not afford. Mortgages
are granted to qualified couples. The trust subsidizes monthly mortgage
payments when necessary, collects payments, and invests whatever capital
is not tied up in mortgages.
To qualify for a mortgage:

1. A couple must have been married at least 1 year but not more
than 10 years.
2. Both spouses must be gainfully employed. At the time of appli-
cation, proof must be submitted of full-time employment for at
least 48 of the previous 52 weeks.
3. The couple meets the LIMT definition of financial need by
meeting one or both of the following requirements:

❑ Installments on a fixed-rate 30-year 90 percent mortgage for a

qualified home exceed 28 percent of their combined gross
income.
❑ They lack sufficient savings to pay 10 percent of the cost of the
qualifying home plus $7,000 (the estimated closing costs).

A mortgage may be issued to qualified couples if

1. The price of the home to be purchased is below the published

median price of homes in its area for the past 12 months.
2. The price of the home to be purchased is less than $150,000.
3. LIMT has sufficient funds to purchase the home.

LIMT calculates weekly its pool of available mortgage funds as follows:

1. Expected annual income from investments is calculated and

divided by 52.

(continued)
 Chapter 6 Method, Process, and Models 171

LIMT—Low-Income Mortgage Trust (continued)

2. Expected annual operating expenses are calculated and divided

by 52.
3. Total income from weekly mortgage payments is calculated.
4. Total of expected payment grants is calculated.
5. Pool amount is then equal to line 1 plus line 3, less line 2
plus line 4.
6. If the cost of homes to be purchased is less than the amount
calculated in line 5, LIMT is deemed to have sufficient funds.
7. Unspent funds each week are invested.

If a mortgage is granted, the couple must

1. File an annual copy of their income tax form to confirm preceding-

year income.
2. Submit copies of pay slips to confirm monthly gross income
each week.
3. Make weekly payments on the mortgage as calculated. Payment
amount may vary from week to week.

Weekly payments are calculated as follows:

1. Capital repayment is 1/1560 of the purchase price of the home.

2. Interest payment is 1/52 of 4 percent of the current mortgage
balance.
3. Escrow payment is 1/52 of the sum of the annual property tax
and the homeowner’s insurance premium.
4. Total mortgage payment is the sum of lines 1 through 3.
5. The couple will pay a maximum of 28 percent of their weekly
gross income. If the mortgage payment calculated exceeds the 28
percent limitation, a payment grant for the amount of the differ-
ence between the mortgage payment and the income limitation
is provided by LIMT. This grant amount is not added to the
outstanding debt.

(continued)
172 Object Thinking

LIMT—Low-Income Mortgage Trust (continued)

Processing requirements are as follows:

1. All data requirements must be updatable.

2. Annual return on investments is updated frequently by the bro-
kerage firm.
3. Expected expenses are updated quarterly.
4. Available funds are computed every week. A report is printed
showing the computation details and results.
5. A listing of all or a selected subset of investments is printed on
demand.
6. A listing of all or a selected subset of mortgages is printed on
demand.
7. Weekly mortgage payment notices (with payment grant
amounts where applicable) are sent to mortgagees following
calculation of the mortgage payment and payment grant
amounts.

The required reports and queries are as follows:

1. List of mortgages by specified price range

2. List of past-due mortgage payments, with indication of which
are fewer than 7 days, 8 to 14 days, 15 to 21 days, 22 to 28 days,
and more than 29 days late
3. List of total grant amounts by month
4. Status of specified account
5. Combined details of mortgage application, mortgage informa-
tion, and payment details

Figure 6-4 shows a semantic net based on the foregoing example.

Remember that this is not intended to be “correct” or complete. A seman-
tic net is a brainstorming tool, and one rule of brainstorming is to record
everything, even if wrong, and save analysis and criticism for later. There
will be nodes and connections on the net that will require clarification and
modification. For example: a couple does not really have an employer—
only individuals do, so it will eventually be necessary to indicate that a

(continued)
 Chapter 6 Method, Process, and Models 173

LIMT—Low-Income Mortgage Trust (continued)

couple consists of individuals and individuals have an employer. Directed

arcs—arrows—would definitely improve the readability of the semantic
net, and you are encouraged to use them in flagrant violation of the syn-
tactic specification for such a net. (Be more daring yet: if the relationship
seems bilateral, use a double-headed arrow.)

makes a weekly

LIMT requires Payment

provides has
Mortgage Investments Payment
Grant

is-for-a is-for-a
Price
must be < or =
completes has-a
Application
Couple
Home Median Price
associates

has-a has-a
Marriage Neighborhood

Employer

Figure 6-4 Semantic net for the LIMT example.

F06MQ04

Object Cubes
An object cube is derived from the CRC card as invented by Beck and Cunningham
and elaborated by Wirfs-Brock (particularly stereotypes). Object cubes offer
single, consistent, metaphor-preserving model objects. The model is used to
aid thinking about decomposition as well as design and some aspects of
implementation. Each of the six sides of the cube, illustrated in Figure 6-5
through Figure 6-10, captures one critical aspect of the conceptualization of
an object.

■ Side 1 (class, responsibilities, collaborations, the classic CRC card)

contains the name of the class to which the object being modeled is
an instance, a list of its responsibilities, and identification of any
required collaborators.
174 Object Thinking

Class: Application Class: InvestmentPortfolio

id self provide total value investments

complete self elements add invesment
validate self validation rule delete investment
return elements return subset of investments
display self elements

Figure 6-5 The classic CRC card view of an object (LIMT example).
F06MQ05

Description: Application Description: InvesmentPortfolio

a form for collection of information about a collection of investments -- Stereotype:

couples and homes --- Stereotype: Form ValuedCollection

Figure 6-6 Object description and stereotype (LIMT example).

F06MQ06

Contracts: Application Contracts: InvestmentPortfolio

(1) public (3) public

(2) private (4) private

Figure 6-7 Contracts: categories of methods or responsibilities (LIMT

F06MQ07

example).

Knowledge Required: Application Knowledge Required: InvestmentPortfolio

id : aString Variable myInvestments : aCollection Variable

myElements : a collection Variable invstmntToAdd : anInvstmnt Argument
myRules : aCollection Variable elmntsRqstd : aSpec Argument
elemntsReqstd : aSpec Argument

Figure 6-8 Knowledge required by an object (LIMT example).

F06MQ08

Draft Message Protocol: Application Draft Message Protocol: InvestmentPortfolio

id add: anInvestment
id: aString delete: anInvestmentId
instantiate return: aSpecification
validate totalValue
rules
rules: aCollection

Figure 6-9 Message protocol.

F06MQ09
 Chapter 6 Method, Process, and Models 175

Events Generated: Application Events Generated: InvestmentPortfolio

funded -- notifies of change in status bankrupt -- portfolio reached zero value

Figure 6-10 Events generated by an object.

■ Side 2 includes a text description of the nature of the objects

represented by the class. Stereotypes (as defined by Wirfs-Brock)
should be included on this side and any notations helpful for
implementation.
■ Side 3 provides a list of named contracts, the most common of
which will be public, private, and protected. Contracts13 are intended
to reveal the intent of the class creator as to who should be able to
send particular messages. Some programming languages enforce this
intent (Java), and others do not (Smalltalk). Each contract lists the
messages (entire signature or just the selector) that should be
included in that contract.
■ Side 4 identifies the discrete pieces of information that an object
will require if it is to fulfill its assigned responsibilities. Each piece of
information is given a descriptive label, a code indicating the source
of the knowledge, and the name of the class that will embody or
contain that knowledge.
■ Side 5 lists all of the messages that an object agrees it will respond
to. Each message listed here will identify the message selector, or
message name; indicate whether any arguments are required (and if
so, the class of each argument); and identify the class of the object
returned in response to the sending of the message.
■ Side 6 names and describes any events (changes in the state of an
object) that other objects might want to be made aware of.

Interaction Diagram
Interaction diagrams are versatile: they’re used to aid discovery as well as to
capture implementation specifications. They model the communication among
a group of objects engaged in a particular task. Most behavioral object thinkers

13. Historically, contracts could be used to aggregate messages by the class of potential senders.
Contracts can also be used to reflect the “method categories” that appear in some IDE browsers
for some languages. Care should be taken in this case to avoid contracts that are language or
programming tool specific.
176 Object Thinking

employ the term scenario for interaction diagrams or a nested set of such
diagrams. Each scenario is bounded by a precondition (what must be true and
what message must be sent to initiate the scenario) and a postcondition (what
is to be returned when the scenario completes normally). Figure 6-11 illustrates
the syntactic components of an interaction diagram. Figure 6-12 uses the syn-
tactic elements of Figure 6-11 to capture a specific conversation—filling in and
validating an application for a mortgage.

Class Class
Message

Object Returned

Message to self Event

Obj Ret
Loop subdiagram
Branch subdiagram
branch subdiagram
(usually error handling)

Figure 6-11 In this set of symbols from an interaction diagram, symbols

F06MQ11

whose labels appear in italics are additions to standard UML.

Use cases, as advocated by Jacobson, are a nested set of scenarios (inter-

action diagrams) whose precondition and postcondition are defined by a single
exchange between a user role and the target system.
A text equivalent of an interaction diagram (see Figure 6-13) is a simple
table with the preconditions and postconditions noted as headers and columns
named Client, Request, Server, Object Returned, and Comments. The Client,
Server, and Object Returned columns contain text labels for objects. The
Request column contains a text description of the request. It will eventually
map to a message sent to the server by the client. The Comments column con-
tains any explanatory text or commentary found to be useful.
A still simpler form (see Figure 6-14) can capture the same information as
a top-level use case diagram: the requests and responses between a user role
and the system. These might also be used to exhaustively list all requests that
might be made by one object of exactly one other object. The model has a
header that identifies the Client object and the Server object. Columns are avail-
able for listing the requests that the client makes, the object returned as a result
of a given request, and comments.
 Chapter 6 Method, Process, and Models 177

DataSource Self Rules

Starter Form EntryField
Interface Evaluat... Collection

Instantiate

nextEntryField

anEntryField
value

mySource

infoSource
value

anObject

validationRules

aRuleCollection
nextRule

aRule

evaluate

self aBoolean

validationRules

aCollection

nextElement

anObject

self

Figure 6-12 This interaction diagram illustrates the conversation required

F06MQ12

to fill in and validate a mortgage application.

Pre: blank form Post: form filled-in

Client Message Receiver Object Returned Comments

Form Next field self An entry field

Entry field Value please Human user A value

Human user value is (x) Entry field self Iterate until all
fields have values

Figure 6-13 Text-only equivalent of the previously modeled interaction

F06MQ13

diagram.
178 Object Thinking

Client: aUserRole Server: anObject

Request Object Returned Comments

Daily report A report Example of message

that might be sent to
the “System Object”
a la a use case.

First element An object An example of a

specific message
sent to an arbitrary
collection object.

Figure 6-14 This text-only binary use case model captures the same
F06MQ14

information as a top-level use case diagram.

These models are useful in defining the high-level responsibilities of an

application in the same way as a traditional context diagram (one system bub-
ble, several external entities, and connecting arcs labeled with highly general-
ized data flows) or a graphical use case model (one system bubble with stick
figures representing user roles, and connecting arcs labeled to reflect general-
ized requests and responses). The intent is to define user/system interactions at
a high-enough level of abstraction to ensure discovery of an exhaustive list.
This type of model might also be used in cases in which an exhaustive list of
potential interactions between two objects is required. I have seen this type of
use in a medical device specification.
It is unlikely that you would use binary use case models to capture all
interactions in a system. The number of models required to do so is n! (n fac-
torial), where n is the number of objects in your system. Even simple systems
would require a huge number of binary use cases to capture all possible inter-
actions. Most developers will reserve the use of this particular model to capturing
interactions with objects or systems external to your immediate development
focus, or to low-level interactions among a subset of objects involved in critical
conversations—wherein every possible interaction must be identified and dealt
with appropriately.

Static Relation Diagram

The static relation diagram is probably the most familiar to any software devel-
oper of all the models presented here. Static relation diagrams depict relation-
ships, relatively static in nature, among a group of classes. Here are three
common examples and their uses:
 Chapter 6 Method, Process, and Models 179

■ Class hierarchy A simple diagram showing a single relationship:

is-a-kind-of. The root (topmost class) of the hierarchy is labeled, by
convention, Object. Many class hierarchy diagrams will be focused
on a subset of an entire domain and might not have Object as root.
Figure 6-15 is an example of a class hierarchy diagram.

Object
Form
DataSource
Interface
Mortgage Valid
Application Collection
Collection Investment

LIMT Mortgage Expense Investment

Collector Collection
Payment Self EntryField ValueHolder
Evaluat...
Mortgage Payment
Employer Collection Collection
Calculation
Element Couple
Person
Bank
Home
Insurance
Agent Assessor

F06MQ15

Figure 6-15 A class hierarchy diagram showing single inheritance only.

■ Gestalt map The most appropriate use of such diagrams is to provide

an overview, or gestalt, of a system or an application while minimizing
the amount of detail. A semantic net is a simple example of a gestalt
map. Gestalt maps, an example of which is shown in Figure 6-16, pro-
vide a convenient shared reference point that reminds developers of the
objects involved in an application and important ways in which they
are related. Some very common types of relationships captured on a
gestalt map include is-a-part-of, is-a-collection-of, is-composed-of,
uses, coordinates, associates, and is-a-kind-of. (These types of rela-
tionships often have their own graphical syntax representation
instead of requiring a text label.) Notice the absence of labels such as
Manages or Controls, relationships that should not exist from an object
thinking point of view. Decorations and annotations added to a gestalt
map enable the capture of constraints or business rules.
180 Object Thinking

accesses
Employer

accesses
Bank

accesses
Assessor
completes
Couple
uses DataSource
Interface

Mortgage contains uses SelfEvaluating

EntryField
Application Rule
accepts
Investment
is-for-a
funds has
LIMT ValueHolder
Mortgage has
Expense
is-for-a Investment Payment Collection
Collection Collection
Home Valued
contains Collection

Collection Payment
uses Calculation
Person Mortgage Form Element
Collection
Object
Insurance
Agent
F06MQ16

Figure 6-16 A gestalt map provides an overview of the objects and their
static relationships.

The more you add special annotations and decorations to the

syntax of a gestalt map, the more likely you are to abuse its intent by
showing too much detail. A gestalt map is not a set of blueprints, using
the analogy of a building; it’s the 3-D model that shows what the
building will look like in its planned setting without the minute details.
In Chapter 10, a very different kind of gestalt map, an “evoca-
tive” map, is introduced. This kind of gestalt map is more appropri-
ate for XP development because it doesn’t fall into the trap of being
exhaustively representational.
 Chapter 6 Method, Process, and Models 181

■ Collective memory map A specialized static relation that shows

how knowledge is distributed among a community of objects and
that provides structural and definitional metadata relevant to that
knowledge. Specifically, it notes the distribution of objects—stored
in instance variables of other objects—whose primary purpose is to
represent some piece of data (strings, numbers, dates, or the like).
Special-purpose structures—for example, valueHolders (a simple
indexed collection, with each element consisting of a label and a
value associated with that label)—are composed entirely of these
data representation objects. Any constraints and relationships that
exist among this subset of objects are noted on this diagram. The
most typical examples of such constraints or relationships are analo-
gous to the key/foreign key relationships found in a standard rela-
tional data model and to class (type) or value domains typically
defined in data dictionaries or schemas. Figure 6-17 shows a collec-
tive memory map.

ClassName.VariableName

initialValue has-a
defaultValue
type (e.g., integer, date, string
domain (set of legal values)
label className.Description

type: valueHolder

composed-of

className.Desc.Label className.Desc.Label .....

F06MQ17

Figure 6-17 A collective memory map, a static view of the “data”

in a system.

Collective memory maps constitute a nod to anyone deeply

concerned with data representations and who miss their entity rela-
tion diagrams. The utility of these diagrams is easily overstated, and
judicious and conservative use is encouraged.

Object State Chart

David Harel revolutionized state modeling with his state charts. Almost all object
modelers now use at least a subset of his notation to capture state-dependent
information about objects. A typical subset of Harel state chart notation is shown
in Figure 6-18.
182 Object Thinking

= State A

trigger (guard)
state 1
action

A = Exclusive - or
B
(either state A or B)

A
Transition via
A B "history" — return
H B
= Concurrency to same substate
(A+C, A+D, B+C, or B+D) as last line
C D C

Figure 6-18 The most commonly used symbols borrowed from Harel.
F06MQ18

Properly defined and designed, most objects will have an exceedingly simple
state. Therefore, the use of this kind of diagram to model object state is seldom
required. Object state charts are extremely useful for modeling state-driven
interactions among objects, such as the interactions that occur in a typical GUI.
I’d like to conclude this chapter by repeating one of the caveats presented
at the beginning: no tool has any value other than to assist in object thinking
and to provide a means for interpersonal communication. If you can model
your objects and your scenarios in your head while engaged in writing code,
and if those mental models are consistent with object thinking, great! No need
to write them down. If you and your colleagues use a visual model on a white-
board as an aid in talking about scenarios and in clarifying your collective
thinking about those scenarios, and you erase the board when you’re done
meeting, also great! If your models are crudely drawn and use only a subset
of the syntax defined here (or a completely different syntax that you and your
colleagues collectively agree upon), still great! Model when you must, what you
must, and only what you must.
Discovery
It is said, “A journey of a thousand miles begins with a single step.”
—Anonymous

This adage is usually quoted as a motivation: “Get started, and the rest will
follow.” The adage misses the potential problem, however, of what happens if
that first step is in the wrong direction. It isn’t necessary to be a fan of chaos
theory to recognize that even the smallest change in initial conditions (the first
step) can have enormous impact on the eventual results.
For instance, suppose you are in Albuquerque, New Mexico. (Bugs Bunny
was famous for making a wrong turn here.) You want to go to Washington,
D.C., following a direct line slightly north of east. Unfortunately, you were
celebrating your coming departure last night, you arose much later than usual,
and, thinking it was morning and the sun was rising in the east, you headed off
in the direction of the sun. You might still get to Washington, D.C. (the Earth is
round, after all), but your journey will be rather tortuous and take an inordinate
amount of time.
Beginnings are just as critical in software development, and initial mis-
takes are a lot more common than one might suppose. A truism of software
development is that the most costly mistakes are made the first day. Two major
factors make beginnings so perilous in software. The first is simply a lack of
knowledge. We know the least about the what, the how, and the why of our
development project at the beginning. This should be so obvious that it need
not be stated, but surprisingly, we insist on making critical decisions about
projects (such as overall design, timelines, and costs) at the very time we are
least prepared to make them.

183
184 Object Thinking

Management usually takes the heat for this mistake. An oft repeated
statement (that I heard in a keynote speech at a conference, but is really a
paraphrase of a famous Santayana quote 1) is, “Insanity is doing the same
thing that failed last time and expecting it to work this time.” Management, it
is said, is guilty of precisely this and therefore should be considered insane.
But it is not insanity at work here—it is culture. If the rationalist philosophers had
been correct and the formalist software engineers had been correct, it would
have been perfectly reasonable to assume that we should be able to define per-
fect specifications and derive working software from those specifications with
mechanical transformations. Computer scientists, software engineers, and ration-
alist western culture in general have been selling management a bill of goods.
The second most common source of error, especially by experienced
developers, is to anticipate how the computer is going to implement your soft-
ware before trying to understand how the software should simulate some part
of the domain in which it is going to be used. This is computer thinking, as
described in earlier chapters. If you start the development process, as is almost
always done, focusing first on what you are going to tell the computer to
do and how it is to do it, you have started your thousand-mile journey with a
180-degree misstep. There may have been a time when computers were so
hard to use, limited in capability, and expensive that it was expedient to use
them and their requirements as a lens through which the rest of the world was
“cut and measured,” but that is clearly not the case today.

Note The argument in favor of computer thinking almost always

revolves around the issue of efficiency or performance. But as the
developers of SIMULA discovered, if you get the modeling and simula-
tion correct, you get a surprising amount of performance as a byprod-
uct. One of the axioms of XP admonishes making things work and
making them right before attempting to make them fast. Even classical
software engineers are adamant about the need to save performance
issues for the very end of development. But it isn’t just performance—
it’s the focus on the machine as the ground for understanding the
software development project that constitutes the wrong first step.

Sometimes it is necessary to foreshadow implementation (programming).

You might need to ask yourself the question, “Can I build this?” Thinking about

1. Original quote from George Santayana: “Fanaticism consists in redoubling your efforts when you have
forgotten your aim.”
 Chapter 7 Discovery 185

implementation can give you a general answer. When you are first thinking
about writing tests and code, you will think about how you are going to imple-
ment a specific story. You will have already done the design, the identification
of objects, visualization of the basic conversations among those objects, and the
assignment of methods to those objects involved in your story—quite possibly
in your head or via discussion with your pair programming partner. Thinking
about the code at this point provides useful feedback. You might discover you
have a bad design, bad language, or code that needs refactoring.
It’s never appropriate to tell yourself, “This is what the code will look like,
so I need an object to hold these parts of the code, and another to hold these
parts, and another to make sure these two do what they are told to do when
they are told to do it,” which is precisely what structured development tempts
you to do.
Perhaps the greatest benefit of object thinking is that of helping you start
off in the right direction. Object thinking does this by emphasizing the need to
understand the domain first.

Domain Understanding
Understanding the domain involves five types of mental discipline reinforced
by object thinking principles and ideas:

■ Extract your understanding of the domain from the domain; use the
metaphor of domain anthropology to guide this effort. The goal of
software development is to create artifacts, with each artifact consist-
ing of an admixture of hardware, software, processes, and interfaces
allowing interaction with other artifacts or with people. Deciding
which artifacts are needed and how they will be required to interact
with existing (and planned) artifacts as well as with people is a result
of understanding the domain—as the domain is understood by its
inhabitants. (See the sidebar in the next section of this chapter,
“Domain Anthropology and Management.”)
■ Decompose the problem space into behavioral objects, and make
sure the behaviors assigned to those objects are consistent with user
expectations. This requires understanding why users make distinc-
tions among objects and the illusions they project on those objects.
User illusions (following Alan Kay) consist of how people recognize
different objects in their world and, having recognized an object,
what assumptions are made about how to interact with that object
and what its responses will be.
186 Object Thinking

■ User illusions should be maintained; your software objects should

not violate them unless you can construct a plausible alternative
story that shows a different set of domain entities interacting in dif-
ferent ways or having different capabilities. Business reengineering
involves exactly this kind of activity—using domain language and
user illusions creatively to craft new stories, some of which might
lead to new software.
■ Decompose your problems (applications) in terms of conversations
among groups of objects. Everything of interest in the domain is cur-
rently accomplished by groups of objects (people and things). Any
artifact you construct must participate in a natural way in these same
groups. Perhaps your artifact is simply replacing an existing object in
the domain with a computer-based simulacrum, in which case it
must know how to respond to and supply some relevant subset of
recognizable and intuitive interaction cues. Perhaps it is an entirely
new object, in which case it will need to be “socialized” to conform
to the existing community.
■ Model your objects as simply as possible consistent with the needs of
the development team (which should include users, à la XP) to com-
municate their collective understanding of the objects and the
domain. Only after you have completed this kind of modeling
should you even begin to think about how to implement any new or
replacement objects your domain understanding leads you to believe
are desirable.

This chapter will discuss how object thinking influences your understand-
ing of the domain, its objects, and the models useful for clarifying and commu-
nicating your understanding. Chapter 8, “Thinking Toward Design,” will outline
how to move toward implementation without abandoning object thinking or
our understanding of the domain.
Domain understanding involves using another metaphor, domain anthropol-
ogy, to provide the mental perspective used when discovering objects and their
interactions. The same perspective is useful for identifying and assigning discrete
services to specific objects. A number of heuristics can be employed to facilitate
object discovery and definition. I will now discuss each of these factors in turn.

Domain Anthropology
The software developer must confront a strange world with the goal of under-
standing that world in its own terms. A close approximation of this task is the
work of a cultural anthropologist confronting a strange (to her or him) society.
Domain anthropology provides a very useful metaphor—providing both
 Chapter 7 Discovery 187

cautions and insights that will guide the application of object thinking to the
early phases (requirements definition, object identification) of development.
Software expertise does not trump domain expertise. The longer a software
developer works in a domain, the more effective her software work will be.
I begin with a caution: anthropologists have their own culture—with its
associated explanations, values, assumptions, and so forth—and they must zeal-
ously guard against the mistake of interpreting everything in the new society in
terms of their own (often unconscious) culture and values. This kind of cultural
relativism should be respected—the software developer does not have a “special”
viewpoint that allows him to judge or alter the culture he is working with.
Software developers, on the other hand, are fair game for evangelistic
efforts. Most developers are steeped in the “way of the computer,” with abun-
dant assumptions about how a computer works and the best ways to make a
computer perform its tricks. You could say that developers have internalized a
culture of computing. The fact that this book—as well as the object and XP
movements in general—is trying to change developer culture does not imply
that developers have a corresponding right to change user or organizational
cultures. Developers, traditional or reformed, must set aside their professional
culture (just as the anthropologist sets aside his or her birth culture) before they
attempt to understand users and user domains. Following this injunction is
inconsistent with the goals and intents of the stereotypical “computer geek”
who has invested a lot of time and energy to think like a computer.

Note To be totally honest, this book is attempting to influence more

than just the culture of software developers. To begin with, I hope it
affects all members of an XP development team, which includes both
users and managers, and that it promotes change of the organiza-
tional culture-at-large in ways that make it more amenable to iterative,
extreme object development. Even more grandiose hopes fill my mind
from time to time—that I will make a small contribution to a movement
that redefines software as a craft performed by human beings on
behalf of other human beings. Beneath the staid exterior of the practi-
tioner-academician beats the heart of a wild-eyed evangelist.

A developer attempting to understand a new application or problem

domain is not unlike an anthropologist confronting a new culture for the first
time. She is surrounded with experts in that domain, but they speak and act in
“strange ways” that she needs to understand. You need to learn from them how
they view and make sense of their world.
188 Object Thinking

Note As an “anthropologist” and an outsider, you do have a perspec-

tive that the domain experts lack. Much of what the domain experts
actually do and think about is invisible to them in the same way that a
human is rarely aware of the culture to which he or she belongs. You will
be able to observe important aspects of the domain that the experts will
fail to report to you or talk about because they are “so obvious” that they
are below the threshold of consciousness. But do not confuse this kind
of objectivity with projections of your own cultural values (computer
knowledge) onto the domain you are supposed to be observing.

When a cultural anthropologist confronts a new culture, the first task (after
allaying suspicion and obtaining some kind of rapport) is to learn the language.
The object developer needs to learn the names of things in the domain that
have been accorded object status by the natives—made distinct from the
domain and from other things in the domain—by virtue of having been given a
name. Naming things is how domain experts (and the indigenous peoples studied
by the cultural anthropologist) naturally decompose their domain.

Note Software developers must also allay suspicion and establish

rapport. An anthropologist does this in two ways, by convincing the
“natives” that he is basically harmless and an amusing curiosity and by
being a source of minor valuables—services such as a ride, small
amounts of money, or goods desired by the people he is studying.
These techniques work reasonably well if the people do not have
extensive experience with anthropologists. They don’t work very well
when there is an adverse history of interactions. Native American peo-
ple, for example, are extremely leery of anthropologists (excepting
sometimes anthropologists that are also Native American) because
they have been exploited too many times in too many ways. The world
of software suffers from this same kind of adverse history. The whole
reason there is so much emphasis on requirements and specifications
is to provide the basis for a contract between two groups that basically
distrust each other. Perhaps the most useful pointers and techniques for
overcoming this mistrust can be found in the work of Larry Constantine
and even the domain-driven design ideas advanced by Eric Evans.
 Chapter 7 Discovery 189

Most books and methods addressing how to do object development recom-

mend that the object discovery process begin with underlining the nouns (names)
in a domain or problem description. While it is true that many of those nouns will
indeed turn out to be viable objects, it is unlikely that any written description will
be sufficiently complete or accurate to meet the needs of domain anthropology.
Cultural anthropologists do not base their understanding of a culture solely on writ-
ten reports by people in that culture. They go and live with the people, observe
them, and talk with them. Domain anthropologists must do the same thing.

Domain Anthropology and Management

Anthropological investigation of a culture is not an easy task. The anthro-
pologist will spend months, if not years, preparing to go to the field. That
preparation includes digesting all that is already known about the target
culture (and other similar cultures) as well as techniques for observation
and information elicitation. Once in the field, the anthropologist will
spend a minimum of a year, optimally two years, living and interacting
with the people he is studying.
A domain anthropologist should do no less! Observation for a com-
plete business cycle (including the annual financial cycle), which includes
the major events and rituals (new employee to company picnic) that occur
in the domain, will yield important information relevant to the design and
success of any new software system.
However, management is unlikely to provide the necessary time for
this kind of activity—at least not as overhead for any single project—if
domain anthropology is seen as a new “phase” of software development
that must be completed “up-front” before any code is written. XP, fortu-
nately, builds domain anthropology into its practices. Early and frequent
releases, short feedback cycles, and open communication all work
together to simulate the participant/observation experience of a domain
anthropologist. As with the anthropologist, these activities begin with
fairly simple day-to-day activities (and the new software to support them)
and evolve, as the developers gain experience with the domain, into more
sophisticated systems. There are other parallels, such as the on-site cus-
tomer playing a role quite similar to that of the cultural informant relied
upon by anthropologists.

Given that management will likely not allow time for a complete ethnogra-
phy, some kind of conversation about the domain with both users and developers
participating is desirable. Constructing a semantic net (see Chapter 6, “Method,
190 Object Thinking

Process, and Models”) is a useful way to gather information about potential objects
and expectations of those objects. While constructing this net, the domain anthro-
pologist must operate, somewhat, as if she were a naive child—constantly asking

“What is this?”

“Why is this considered to be a ______?”

“How do you distinguish between this and that?”

“Why?”
The answers to these questions not only reveal potential objects but also start
to reveal the expectations that domain experts have of those objects. Begin the
object discovery via the semantic net technique by asking for one or two sen-
tences about the domain in general, such as two sentences that describe what your
company/department/team does. Select a couple of nouns and verb phrases, and
write the first elements of a semantic net (see Figure 7-1) on a whiteboard. Ask
everyone present to brainstorm new elements (circles, nouns) and new connections
(arcs, verb phrases). This exercise should be a stream of consciousness—
brainstorming—session, with analytical discussion to follow later.

Completes
Couple Application
Mortgage
Trust
Has_A
Is_For
Funds

Legal MTGE
Marriage
Is_An
Investments
Secured_By
Has_A

Home
Marriage Is_In_A
Date

?????? Neighborhood

Today's
Date

Figure 7-1 A partial semantic net for a subsidized mortgage company.

F07MQ01
 Chapter 7 Discovery 191

It usually takes but a short period of time, five to ten minutes, to produce a
fairly large diagram. The circled nouns provide a rich set of potential objects, and
the relationships provide an equally rich set of potential responsibilities (expec-
tations of those objects). This semantic net can then be placed on the wall as a
common reference point as the discussion proceeds to more detailed levels.

Subsidized Mortgage Company

The partial semantic net in Figure 7-1 is based on the following conversation
about a company that provides subsidized mortgages to qualified couples.
Please give me a brief description of your company.
Our company provides mortgages to qualified couples, subsidizes
monthly mortgage payments when necessary, collects payments, invests
whatever capital is not tied up in mortgages, and otherwise manages its
assets. (This yields an initial net consisting of bubbles for mortgage, couple,
payment, investment, and asset. Initial arcs include investment is_an asset,
couple receives mortgage.)
I assume the mortgage is for a home?
Yes, a home that must have a purchase price less than or equal to the
median price of all homes in its neighborhood. Oh, and the home cannot
have a purchase price greater than $100,000.
What qualifies a couple for a mortgage?
Couples must have been married for at least 1 year, but not more than
10, and be gainfully employed. They will not have the required 10 percent
down payment required by other lenders, and the monthly payments on a
$90,000 mortgage will exceed 28 percent of their joint gross income.
What do you mean by gainful employment?
At the time of application, proof must be submitted of full-time
employment for at least 48 of the previous 52 weeks.
We also have to have funds available that are sufficient to cover the
mortgage.
How do you know if you have funds available?
We calculate our expected income from investments and divide by 52
(we calculate available funds weekly); we figure our annual operating
expenses and divide by 52; we total weekly income from mortgage pay-
ments (oh yes, we also collect mortgage payments weekly); and we calculate
any grants we expect to give out this week. Expected annual income from
investments is calculated and divided by 52 and added to the pool of funds.
We then compare the cost of the mortgage or mortgages applied for with

(continued)
192 Object Thinking

Subsidized Mortgage Company (continued)

the amount of money in the pool and fund all we can. Any leftover money
is invested.
People have to apply for the mortgage? (Note the naive question.)
Of course they fill out an application, and when the information on
the application is verified, the application is deemed fundable. If we have
money, we grant the mortgage and the application is given funded status.
If we do not have the money this week, the application stays fundable
until it is funded or some other event causes it to be canceled.

One reason for reducing the object discovery process to simple questions
and observations is to eliminate preconceptions. If a cultural anthropologist
were to begin an analysis of kinship with the preconception that everyone lives
in a nuclear family (mother, father, and 2.5 children) and that everyone practices
monogamy, her analysis of human kinship patterns would be so incomplete and
distorted that it would be worthless.
An analogous result will occur if the domain anthropologist begins her
task with preconceptions about data, functions, processes, how things must be
implemented in COBOL, or how computers work. These preconceptions about
implementation will blind the domain anthropologist to what is going on in the
domain and what is required of the new software artifact. For instance, a soft-
ware expert is likely to see any kind of form as an ordered collection of static
text objects and entry field objects because that’s the way that a form will be
implemented. If this knowledge disposes the expert to ask users only those
questions dealing with choices of text and entry fields and the values that can
be placed in an entry field, the expert will miss potentially important behavioral
requirements. Behavioral requirements for a form might include identifying
itself, modifying itself, presenting itself (with the implication of multiple views),
completing itself (making sure every entry field contains appropriate values),
and validating itself. These are behaviors that will not be discovered by asking
about contents.
The initial focus of domain anthropology is on identifying stories and
some potential objects in the domain. Stories reveal both potential objects and
some of the “social relationships” among them. Human cultures have rules for
interaction among their members. People engage in a complex web of mutual
obligations that change with time and circumstance. People participate in
relationships, some fixed (such as biological parentage) and others temporary
(such as employer-employee). Similar social relationships exist among objects,
and they too must be discovered.
 Chapter 7 Discovery 193

Note Another warning: We have “trained” users over the years to

anticipate how software developers think. It is not unlikely, therefore,
that they, at least initially, will try to tell you what they think you want to
hear. They will couch their responses to your questions in terms of
data, attributes, algorithms, and functions. When this happens, they
must be questioned further to discover the behavior involved.
The metaphor of an object as a person is important for developers
to begin to understand objects. It is equally useful for domain experts
and should be discussed with them preliminary to any question-and-
answer session. Both the developer and the domain expert can then
use a common language in their discovery dialog. For example:
“Well, there are customers.” (User offers a potential object.)
“What do you expect of a customer?”
“It should tell me its name, or identify itself.”
“What else?”
“It has a unique identification number.”
“Do you want it to tell you that number?”
“Yes.”
“Then we can say it must report or tell you its unique identifica-
tion number.”
“Yes, and it should tell me its address, phone number, age, gen-
der, birth date ...”
“Wait a second, you mean it should be able to describe itself and
tell you how to contact it?”
“Yes, yes, exactly.”

Details, and in some cases discovery, of some relationships will occur only
as attention shifts to defining tests and writing code (and the implicit design that
occurs simultaneously). The domain anthropologist (embodied in the on-site
customer working with technical developers) focuses on how and why objects
interact and, in some cases, why they do not interact—in other words, what rules
prevent a specific object from certain types of interactions with other objects.
How objects interact is mostly a matter of recording actual and possible
(desirable) conversations among objects. Why is mostly answered by seeking
the task or tasks that motivated the conversation. There are different ways to
organize the work required to solicit all the conversations and tasks relevant to our
understanding of the domain and of the objects we might have to build to operate
in that domain.
194 Object Thinking

One way to organize this work—fairly traditional in a software engineer-

ing sense—would be to isolate a subset of the objects in the domain, draw an
arbitrary boundary around them, and label the collective the NewSystem object.
You could then ask each remaining object in the domain to identify the services
it might require of the NewSystem object. The set of nonduplicated services con-
stitutes the responsibilities/behaviors of NewSystem. These, in turn, will need to
be examined to expose the actual multiobject conversations required to provide
the identified service. Conversations might have subconversations. Eventually
this process will yield a set of nested conversations, each nested thread being a
use case as that concept was developed by Ivar Jacobson and incorporated into
UML. Alternative, and probably better, ways to approach this work can be
found by reading Alistair Cockburn’s work on use cases and Larry Constantine’s
discussion of “essential form.”
Each threaded conversation can be modeled, using UML syntax, with a
high-level diagram (similar to the context diagram of structured development)
that represents each service consumer (user role) as a labeled stick figure and
the NewSystem object as a labeled circle. (See Figure 7-2.) The making of a
request by a service consumer defines a precondition for some conversation
among objects within the NewSystem circle. The return of an object to the ser-
vice consumer constitutes a postcondition for that same conversation. The con-
versation itself can be modeled with an interaction diagram. (The syntax used
to construct interaction diagrams is covered in Chapter 6, “Method, Process, and
Models.”) Interaction diagrams can be nested, although UML graphical syntax
provides no way to represent the presence or absence of a subscenario. The
extended UML syntax shown in Chapter 6 does provide for branching and
looping subscenario notation for the interaction diagram.

Pre-Condition: Request 'x'

Class Class Class

1 2 3
Request 'x'
Request 'x'

User Role
Response 'A'
The
Request System
Response 'A'

Response
User Role

Figure 7-2 A UML-style, top-level (analogous to a context diagram) par-

F07MQ02

tial use case.

 Chapter 7 Discovery 195

Note Use whichever syntax (graphical or textual) you prefer.

Remember, object philosophy questions the utility of any representa-
tion or model beyond its immediate assistance to those involved in
making the model. Don’t get caught up in trying to make your model
“complete,” “accurate,” or “true.”

This same information about the objects in a domain and their interaction
can be modeled textually. The system request table (Figure 7-3) replaces the
stick figure diagram of UML, and the conversation table replaces the graphical
interaction diagram (Figure 7-4).

Figure 7-3 In this system request table, the service requester is identi-
F07MQ03

fied at the top. (NewSystem is assumed to be the service provider.) Each

line identifies the service requested: the message sent to NewSystem,
the object returned as a result of that request, and notes for explanations
if needed. Each line of the table constitutes a precondition/postcondition
for a conversation.
196 Object Thinking

Figure 7-4 In this conversation table, preconditions and postconditions

F07MQ04

are noted at the top of table. Each line of the table shows who is making
a request, the nature of the request (message), the service provider, the
object returned, and a comment if necessary. The conversational order is
assumed to be top down.

There is nothing necessarily wrong—or non-object-thinking—with the

systematic approach just outlined. To be systematic, however, it presumes a
concerted effort to identify all use cases up front and is based on defining the
requirements for NewSystem. An alternative approach, one much more consis-
tent with XP ideas and techniques, would be to define stories one object (or a
small number of objects) at a time. All the same elements of discovery are present
that would be in a systematic approach—it’s simply a philosophic preference to
do things in the small.
We begin, this time, by looking at the semantic net constructed earlier and
noting the relationship (completes) between couple and application. We then
write a simple story about this relationship:

A couple requests an application for a mortgage and fills in all

necessary blanks.

This story gets expanded into a sequence of interactions:

1. The couple glances at the application to make sure it is the right one
(implicitly asking the application to identify itself).
 Chapter 7 Discovery 197

2. The couple looks for the first blank spot (implicitly asking the appli-
cation for the next space needed to be filled in).
3. The couple asks what information needs to be placed in that spot
(implicitly asking the blank space for its label, on the assumption
that the label will describe the information needed).
4. The couple enters a value in the blank spot.
5. Steps 2 through 4 are repeated (next blank spot replacing first blank
spot in the description of step 2) until there are no more blank spots.

This conversation can be graphically depicted with an interaction diagram,

as shown in Figure 7-5.

Figure 7-5 This interaction diagram depicts the user-driven conversa-

F07MQ05

tion required to fill in a form.

This conversation assumes an active couple object and a passive form

object. In one sense, this reflects reality: a paper form is pretty passive, and
a couple is made up of live human beings capable of action. Nevertheless,
it isn’t as reflective of object thinking as it might be. Consider the alternative
story card:

A couple requests an application. The application, the couple,

and entryField objects collaborate to make sure every entryField
needing a value contains one.
198 Object Thinking

This leads to a conversation with the following steps (graphically depicted

in Figure 7-6):

1. The couple requests an application from SubsidizedMortgageCompany.

(Our conversation reveals a new object capable of providing an
application.)
2. The application presents itself to the couple. (Implicit identity verifi-
cation plus ability of the object to display itself.)
3. The application identifies the next entryField (or fields, since both a
paper form and a GUI-based computer screen allow multiple entry-
Field objects to request input simultaneously) requiring content by ask-
ing itself for the next field (or fields) and then notifies that entryField
(those entryField objects) that it/they now has/have the opportunity to
obtain content.
4. The entryField asks the couple for input. (Implicit request, mani-
fested by asking a blinkingCursor to display itself in the area of the
entryField or by highlighting itself in some other way.)
5. The couple provides a value.
6. The entryField notifies the application that it is satisfied.
If it is not satisfied, an exception condition exists, and a story
describing how to handle that case must be defined.
7. Steps 3 through 6 are repeated until the supply of entryField objects
associated with the application is exhausted.

Form Entry Couple

Field (any value source)

instantiate yourself
next
entry field
a field
instantiate yourself
value please

a value
validate
self
self

Figure 7-6 This interaction diagram depicts the alternative scenario for
F07MQ06

filling in an application.
 Chapter 7 Discovery 199

Differences in the conversation are minimal, but they reflect the behavioral
parity of all objects—that every object, including humans filling the role of infor-
mation provider, are defined in terms of the services they can provide others.
Each object is an active entity in control of its own destiny. The alternative way
of thinking about the conversation also reveals objects that might not have been
discovered or thought about in the original model of the conversation.
More important, the thinking behind the alternative conversational model is
more general—more abstract—and therefore applicable regardless of the means
of implementing the application. Behavioral expectations of the application and its
parts are consistent whether implemented with paper or as a software application.
True, the software versions of the objects have more interesting or dynamic ways
of exhibiting behavior. An example is the entryField object using the services of a
blinkingCursor instead of simply displaying itself as a blank space.
The alternative model is also more consistent with user expectations of the
objects involved, which is the real goal of domain anthropology. If we are to
build digital versions of forms and entry fields, they should reflect the expecta-
tions of their naturally occurring counterparts. For example:

■ We expect application forms (any form, for that matter) to “present”

their constituent parts (labels and entry fields) to us in an organized and
reasonably clustered way. In the digital manifestation of the form, this
is accomplished through interaction with a cursor or, more crudely, by
presenting the entry fields to the user one at a time. In the paper imple-
mentation, the form still presents the elements to us in order but does
so implicitly—taking advantage of cultural conventions.2
■ We expect entry fields to ask us to enter a value. The digital version
does so by using a blinking cursor assistant object to remind us that it
is waiting and refusing to go away until it is satisfied. The paper ver-
sion does so by nagging us simply by remaining blank. Nature abhors
a vacuum no less than a human abhors a blank spot on a form.

The value of applying object thinking to conversation discovery and model-

ing will become more evident as we consider responsibility assignment and
discovery in more detail. Here’s a preview: allowing entryField objects and form
objects to validate themselves in collaboration with another object (a rule object).
A paper-based entry field seeks validation by displaying its content to a human and
asking for confirmation or replacement of its displayed contents. A digital version
could do the same thing by asking a rule or rules to evaluate themselves (to true or
false) using the value held by the entry field as one of the rule variables.

2. Reflects U.S. and Western culture. Other cultures use right-to-left and down before left-to-right
conventions.
200 Object Thinking

A reader note earlier in this chapter suggested that a domain anthropologist

brings to bear a certain sensitivity or objectivity by virtue of being an outsider.
Seeing the implicit behavior of forms and entry fields, just noted in immediately
preceding paragraphs, is an example of applying this kind of objectivity. It
would not occur to a typical user to volunteer how they interact with forms and
entry fields because that interaction is based on cultural conventions that are
nonconsciously applied. The conventions are nevertheless real and need to be
identified and exploited in order to develop and design the most robust objects
and applications.

Object Definition
Semantic nets and stories (scenarios, use cases, interaction diagrams) provide
input, but the most critical aspect of discovery is object definition. Each object
must be identified in terms of its actual or intended use in the problem space—
the domain. The metaphor of domain anthropology continues to shape our
thinking about each individual object and what we expect of it in terms of services
(behaviors, responsibilities).
Each circle on the semantic net is a potential object. We can select any one
of them and attempt to identify all of the services such an object might provide in
its domain—in its domain, not just the application or specific problem where we
first encountered this object. It is true that object thinking is biased in that it expects
most objects to have limited and simple behaviors—which makes it possible to
model them in terms of an entire domain rather than an application. This bias is
supported by experience, but it is an inductive conclusion, and the exception or
exceptions may yet be found. Typically we would use the first two sides of the
object cube model to record our thoughts about the object and its behaviors.
Our goal during discovery is object definition, not object specification.
Definition means we want to capture how the object is defined by those using
it, what they think of that object, what they expect of it when they use it, and
the extent to which it is similar to and different from other objects in the
domain. Specification will come later (maybe 30 seconds later if you are doing
XP and working on a single object), when we allow ourselves to consider how
this object might be implemented (simulated) in software. Specification will
involve making some design decisions and capturing the information necessary
for someone to actually build the software version of the object.
Object definition involves capturing three bits of information: a short
prose description of the object (in the words of a domain user), an enumeration
of the services it is expected to provide (and, if other objects are used as helpers
for any given service, the kind of object to be used), and, when appropriate, a
stereotype—another object or other objects that the one we are working with
resembles in terms of similarity of services provided. Wirfs-Brock and McKean
 Chapter 7 Discovery 201

identify several stereotypes, including information holder, structurer, service

provider, coordinator, controller, and interfacer. Peter Coad has also written
about a wide variety of stereotypes. Ivar Jacobson uses the idea of stereotypes
with emphasis on model (service providers), views, and controllers.
Side 2 of the object cube is used to record the description and any stereo-
type; side 1 is used to record the list of responsibilities and collaborators. To this
point, we are engaging in exactly the same thinking and recording process
advocated by Beck and Cunningham and their CRC cards. It would not be
unusual to actually use 3 × 5 index cards for discovery because they are cheap,
easy to modify, and shareable (you can pass them around); and they provide a
good visual reminder that well-defined objects have limited responsibilities.
Where to start? Pick an object or two from the stories selected by the cus-
tomer for the first development iteration. This will provide a starting point. For
example, Application and entryField. We ask the user (domain expert) for a def-
inition, and the user responds, “An application is a kind of form used to collect
specific facts about the people applying for a mortgage and the home they want
to purchase.” We record this information on side 2 of the object cube (or on the
back of the 3 × 5 card). The definition, conveniently, provides us with insight
into a stereotype for the application object—it’s a “kind of form.” So we record
that fact on side 2 as well. (See Figure 7-7.)

Application [side 2]
A kind of form used to collect facts
necessary to make a leading decision

stereotype:
a Form
a Collection
(when computerized) an Interactive Form

Entry Field [side 2]

A labeled space where data/facts can
be entered and recorded. The label id's
what fact is wanted. Fields must be
validated.

stereotype:
an EntryField

Figure 7-7 Side 2 of the object cube for the Application and entryField
F07MQ07

objects contains a simple record of the description and stereotype

obtained from the domain expert (user).
202 Object Thinking

We do the same thing for the entryField object, obtaining the information
shown in Figure 7-7.
We now turn our attention to the services we expect from our Application
and entryField objects. We have several choices of how to think about the ser-
vices. We can look at the arcs connecting the application circle to others on the
semantic net to see whether they suggest services expected of the application.
We can go over the story about filling out the application to see what was
expected of the application in that story. Either of these options is likely to
expose services that our application must satisfy, but those services might be
couched in specific language reflective of the limited context in which we dis-
cover the service rather than the domain as a whole. We will take care to gen-
eralize any such discovery so that the service reflects the needs of all of the
objects that might use our application.
A third way to approach the enumeration of services is simply to ask the
question, “I am an application; what services can I provide to others?” To help
us answer this question, we can look at an actual physical example of the appli-
cation form, and we can begin to collect those stories involving the application
that have been provided to guide our immediate development activities and
phrase the question in the context of each of those stories. “In this story, what
services do the other objects expect of me?” As your project evolves and as the
application object is discovered in other projects with other stories, you revisit
your thinking about the essential services that your object is to provide.
We notice that the application has a funny string of characters at the bottom
(something like AF001rev10/03), and we also notice a line of print at the top in
larger font size than the other contents. We surmise that these character strings
identify the form in some way. We also recall that the stories told about the
application implied that users of the application looked at it to confirm that they
had the right form. From this we conclude that the form has a responsibility to
identify itself. We record this information on side 1.
Visual inspection reveals several examples of short text strings and several
examples of blank spaces. We might give these things names—textString and
entryField, respectively—and then create a new object cube with those names
on them. We will want to explore those objects later to determine what their
responsibilities might be.
Our attention is on the application, however, so we wonder whether the
application has any responsibility in connection with these newly discovered
objects. It’s obvious that the application holds these elements in an organized
manner. We might arrive at a responsibility that the application holds elements.
We reject the phrasing holds elements because that does not describe a service;
instead, it describes something the application must do to provide a service. If
we say holds objects, we are implying an implementation decision: how an
object is to do something rather than what it is to do. A bit of thought, and we
 Chapter 7 Discovery 203

come up with the responsibility provide access to elements. Implicit within this
responsibility are variations such as provide access to a subset of elements and
provide sequential access to elements. We might record these variations on side 1
so that we don’t forget them, even though we aren’t sure they are appropriate
responsibilities.

Note Abstraction, generalization, classification, and essentialism are

three quite different but apparently similar concepts. Abstraction is the
act of separating characteristics into the relevant and irrelevant to facil-
itate focusing on the relevant without distraction or undue complexity.
Generalization also involves separation but more in terms of shared
and not shared characteristics. Classification involves (depending on
the linguistic theory you prefer) either separation based on conform-
ance to a set definition or similarity to a prototype—the separation
being in the class or not. Essentialism is an attempt to separate acci-
dent and essence—to identify the characteristic that is essential to
being considered an instance of some class.
Unfortunately, abstraction and generalization tend to be used as
synonyms for all four terms as well as for each other. I am as guilty of
this as anyone else—hence this note and its attempt to deal with an
apparent contradiction between object thinking and XP.
Kent Beck says, “Keep things simple by not providing abstractions
until the abstractions provide simplicity,” one expression of his consis-
tent warning against “premature” abstractions. He, and most followers of
XP practice, believe that abstraction should arise from refactoring rather
than from activities typically associated with the analysis and design
phases of traditional development.
Object thinking suggests you should take care to generalize
responsibilities so that they can be used in any context. This is not
abstraction of the type chastised by XP. It is much closer to essentialism—
finding the specific traits or characteristics that make a thing a thing.
Sometimes this is as simple as removing adjectives from a service
description (method name). At other times, it is subtler and a bit closer to
generalization: instead of “Provide unique identification number” as a
responsibility that reflects how an object identifies itself in a context, use
“Identify self.” A significant part of object thinking is devoted to providing
the philosophy and metaphors and definitions required to accomplish
this paring of things to their essential nature. I believe that the apparent
contradiction between object thinking and XP is just that—apparent—
and derives from nuanced use of common terms.
204 Object Thinking

We can see the application, so we know it can display itself. We would

record “display itself” on side 1. It’s necessary to think a bit about what’s
involved in displaying the application because more than one object is visible:
the application, of course, but also the textString and entryField objects. It’s
inappropriate for one object to assume responsibility for the actions or behav-
iors of others. The text strings and entry fields must display themselves. The
application might be the recipient of the display request, but it must be careful
to delegate some of the display behavior to its elements. We can capture this
information by recording that myElements is the collaborator used by the appli-
cation to make sure all objects are displayed.

Note As you tackle more stories involving forms, you’ll discover

additional abstractions or recursive use of the same abstractions. For
example, a complex form might have sections or pages that are them-
selves forms with their own elements—an example of recursive orga-
nization. This kind of organization reflects the composite pattern.
Some forms will have check boxes or other means of collecting infor-
mation. At first, these appear to be something other than an entry-
Field, which in our discussion so far appears to be a box for collecting
a string. A bit of reflection reveals that a check box—and almost any
other widget you might find of use in simulating a form—is, in terms of
its behavior, just an entryField. Different widgets have different
appearances, and thinking about that fact might lead you to positing a
view object, charged with providing a visual presentation of an object.
An object that exists in different view contexts would have multiple
view objects at its disposal, one for each context. An apparently simple
responsibility of the form object—display itself—turns out to be quite
complex when we start to think about writing the code that enables
that responsibility. But we want simple code as well as simple respon-
sibilities. To obtain simple code, we must refactor—which means think-
ing about how we can break up the code and distribute it to other
objects. In this case, part of the responsibility gets delegated to the
elements that the form comprises and still further to objects that spe-
cialize in visual representation tasks—view objects.

Usually we want to record only one collaborator on a line. To accomplish

this heuristic goal, we might stereotype both textString and entryField as dis-
playable objects. We can then record displayableObject as the collaborator. It
would be nice to be a bit more specific about instances of displayableObject
that we are collaborating with, so we might refine our label (which identifies a
new class to be thought about) to displayableElement instead. This implies that
 Chapter 7 Discovery 205

any elements the application contains that are displayable are the ones we
collaborate with to complete the application’s display responsibility.
The fact that the application is a container of other objects also suggests to
us a second stereotype for application. It has behaviors similar to a collection—
another type of container object. We should add this observation to side 2 of
our card.
One more responsibility needs to be considered. In one of our stories, it was
suggested that the application form needs to be verified or validated. The informa-
tion entered on the form—the information contained within the entryField
objects—must be confirmed. This would suggest a responsibility validate applica-
tion form. But who should this responsibility be assigned to? According to the pre-
cepts of object thinking, the application must be responsible for validating itself!
So we add one more responsibility to side 1 of the object cube, validate self.
Exactly how this is done we do not know. But a bit of reflection indicates that
part of the validation is done on the values stored in each entryField. The appli-
cation should not be validating the entryField; the entryField must be responsible
for that itself. Just as was done with the display responsibility, the application will
delegate to each entryField its share of the validation workload. We list entryField
as a collaborator for the validate self responsibility of the application.
Alas, our thinking may be leading us astray a bit in our consideration of
validation. Perhaps it is not one task but two similar tasks, one of which
belongs to the application and the other to the entryField. To be sure we are
doing the right thing, we need to ask what is meant by validation as that term
is used in the domain. So we might ask a user, “How is an application vali-
dated?” The user then provides a simple story about validation:
“Well, the data in each field is checked to see whether it’s the
right kind: a name, a date, a dollar figure, things like that.
Then we verify that the value is accurate and permissible. By
permissible we mean things such as the number entered in age
has to be between 25 and 40, or the number of years married
must be between 2 and 10. By accurate, we mean that the
income figure is verified with the employer. Also, the bank
balances, things like that. Finally, we check the application for
consistency. Say they enter a city name and a postal code. We
check to see that the postal code matches the city.”
Thinking about the story suggests that the application, as well as each
entryField, is validated by applying a rule or rules to the value stored in an
entryField. In the case of the application, the rule might have to look at values
stored in more than one entryField.
Our first attempt to implement this story might utilize a lot of If statements and
the hard-coding of a lot of values in the conditional checks of those If statements.
206 Object Thinking

There is significant potential for inappropriate coupling of our objects—for

example, if we had a “validate” method for a form with code that looked like
the following:
Validate
If form.city concatenated with form.zip is found in zipcodeDirectory return
"TRUE" Else return "FALSE."

If the names of fields on the form change or if the rule changes in some fash-
ion, we will have to fix the code. There is insufficient flexibility and too much reli-
ance on conditional code to pass the XP “smell test.” Applying the object thinking
principle—everything is an object—in thinking about how to refactor our code
should lead to the discovery of another kind of object—a rule object. A rule object
is given the responsibility to evaluate itself to some value (either the TRUE or
FALSE of this example or any other value). We can then provide the form and each
of its entryField objects with a rule (or rules) that it can use to effect its own vali-
dation. So we change both object cubes to reflect the newly identified collabora-
tor. We would also make a new object cube for the validationRule object.
When we are done, we probably have the object cube—sides 1 and 2—for
Application and for entryField completed as shown in Figure 7-8. Once we com-
plete sides 1 and 2 of an object cube for every object identified in our semantic
net, and once we have assured ourselves that we can tell any story the user can tell
us about using those objects, we are done. The overall result for the subsidized
mortgage company will be similar to Figure 7-9 (side 1) and Figure 7-10 (side 2).

Class: Application Class: EntryField

instantiate self instantiate self
EntryField validate self
validate self display self
display self update validation rules
EntryField
reorganize self
update validation rules

Description: Application Description: EntryField

A kind of form that collects, saves, and A labeled object that collects, validates, and
validates information required to approve a saves a particular bit of information.
mortgage. Stereotype: a form, an interactive The label identifies the type of information
form, a collection. expected. Stereotype: an EntryField.

Figure 7-8 Sides 1 and 2 of the Application and entryField object cubes.
F07MQ08

"(Indented lines on side 1 of Class: Application are collaborators for the

immediately preceeding responsibility.)"
 F07MQ09
Class: Object Class: LIMT Class: Payment
Object Object Class: DataSourceInterface
display self Object
fund mortgages calculate self
id self invest surplus return calculation amounts return parsed value
persist return expense total indicate value change
grant access return surplus total implement collection behavior
describe self return income total
dispatch state changes

Class: Form Class: Collection Class: Mortgage Class: Investment

Object Object Object Object Class: Employer

of the object cubes).

fill self in add element return application id return periodic income
EntryField delete element return couple return current value return weeks employee worked
validate self return subset of elements return home return base value return wages employee earned
return status return current gain/loss
indicate terms return object invested in

Class: MortgageApplication
Class: ValuedCollection Class: ValueHolder
Form, Object
Object Object Class: Bank Class: InsuranceAgent
return funding status
return cumulative value return keys
implement collection behavior return value at key return savings amount provide insurance premium
implement collection behavior return checking balance
return customer rating
return accounts for customer

Class: EntryField
Object Class: InvestmentCollection
Object, ValuedCollection Class: CalculationElement
Class: Home Class: Person
obtain value
validate self return periodic income
calculate self
return validation rules return sale price provide employer
return held class name return area provide bank
return legal description

Class: SelfEvaluatingRule
Object
evaluate self Class: Assessor Class: Couple
Payment Collection Mortgage Collection

return tax assessment return bank

Chapter 7

return median price for given area provide marriage date

provide person

Figure 7-9 Objects identified in the subsidized mortgage example (side 1

Expense Collection
Discovery
207
 208

F07MQ10
Description: Object Description: LIMT Description: Payment
Description: DataSourceInterface
root of class hierarchy this object represents the mortgage a document identifying a periodic
trust and functions as the "global" obligation to transfer funds and a representation and interface to an
object in many application programs. documents that transfer when it occurs. outside object from which objects in
our applications require information.
Object Thinking

Description: Form Description: Collection Description: Mortgage Description: Investment

Description: Employer
a kind of collection that contains a general class that adds, deletes, document that contains information a record of money invested and
strings and entry fields - can be organizes, returns all or part of its identifying the mortgage obligation expected returns from that investment an instance of data source interface
instantiated to any form merely by contents and all of its legal requirements. It including periodic dividends or
adding, deleting and rearranging its has a decreasing value as payments payments
elements are received

(side 2 of the object cubes).

Description: MortgageApplication
Description: ValuedCollection Description: ValueHolder
Description: Bank Description: InsuranceAgent
a kind of form that accepts and
a kind of collection that has a a kind of indexed collection that can
records information required to
cumulative value only contain "primitive" data type an instance of data source interface an instance of data source interface
make a mortgage funding decision
objects

Description: EntryField
Description: InvestmentCollection
Description: CalculationElement
Description: Home Description: Person
accepts and validates an entered bit
of information, can change its an instance of a valued collection
a kind of self evaluating rule for
validation criteria a representation of facts about a a representation of a human being
calculating part of a payment
dwelling

Description: SelfEvaluatingRule
Description: PaymentCollection
Description: MortgageCollection Description: Assessor Description: Couple
a rule that can reformulate itself by
adding and deleting elements and an instance of valued collection
resolve itself to a final value which <description unspecified> an instance of data source interface an instance of data source interface
it can report back

Figure 7-10 Objects identified in the subsidized mortgage example

Description: ExpenseCollection

<description unspecified>
 Chapter 7 Discovery 209

As you look at the figures, you’ll see objects that we have yet to discuss in
the text—objects that have come to light as other developers worked on their
stories. As the team works, objects will be proposed and developed by different
pairs, and you might not be aware of much more than the names of the objects
until you have the opportunity to work on a story involving those objects.
Object discovery provides us with a set of named objects and a set of
responsibilities assigned to each one. We might also have a short prose descrip-
tion of the object and a stereotype. In one sense, we are ready to code the
objects we have discovered, and that would be the expected activity of the XP
programming pair (coding tests first, of course).
Object thinking can still help you as you think about the code and start to
make implementation decisions. For example, we can use the stereotype to sug-
gest an existing object that might be modified or employed to help fulfill the
responsibilities of the one we are working on. This kind of implementation deci-
sion will give rise to a taxonomy of classes—a class hierarchy. Such a hierarchy
has several benefits for programmers: providing already written and tested code
that we can employ for our own ends and, because of the taxonomic organization,
a kind of index we can use to look for code that might be useful to us.
Chapter 8 will introduce some additional aspects of objects that we can
think about to facilitate our implementation—aspects that might be characterized
as design considerations. In Chapter 10, we’ll look at class hierarchies (taxono-
mies) in a bit more detail.

Another Example
Some Objects from an Air Traffic Control (ATC) System
Imagine we are working on modeling the world of air traffic control. We
will have identified a number of objects in that realm, including passen-
gers, airplanes, control towers, flight routes, controlled airspaces, and
many more. We will also have captured a number of stories about how
objects interact in this world. The following is a brief illustration of how a
few objects might have been modeled in such an example. We will focus
on the airplane and some objects that come to be identified as we con-
sider the airplane and its responsibilities. We will use the story of how an
airplane fulfills one of its responsibilities—reporting its location as the
background—for our process of discovery.
Figure 7-11 includes a depiction of side 1 of the object cube for an air-
plane. By identify self, we mean that the plane can provide its registration
number. Describe self entails giving a requester of that service an object

(continued)
210 Object Thinking

Another Example (continued)

containing the characteristics of the airplane (type, manufacturer, seating

capacity, and so on). Report location means the airplane will provide a loca-
tion object containing all the pertinent information for a location (altitude,
latitude, longitude, vector, and so forth). Move to new location means the
airplane will actually relocate itself in space.
The only one of these responsibilities that looks complicated to code
is move to new location. A fair amount of calculation is required to deter-
mine the combination of acceleration and attitude that will be required to
get from “here” to “there,” and then directives need to be issued to the
objects capable of effectuating those changes.
There is another, subtler, complexity in the plane’s ability to report
its current location. Although the explicit responsibility is to provide a cur-
rent location, there is an implicit responsibility to report the correct current
location. A location is a composite of several values—all of which are
dynamic—and the only way that the plane can assure itself of the correct-
ness of the location it is providing is to have one constructed in response
to the request it receives from the control tower.
We probably do not want to add a responsibility to the airplane of
the sort, construct location. Instead we acknowledge the need for con-
struction in the original report current location responsibility by adding a
collaborator to that line. We then expect the collaborator to assume the
task of construction. Object thinking mandates that objects do things to
themselves, so the obvious collaborator is a location object.
The airplane can ask its collaborator, the aLocation object, to instantiate
itself (obtain and remember all the appropriate values for altitude, latitude,
longitude, airspeed, and so forth) immediately upon receiving the control
tower’s request for a location. The location object then interrogates the
necessary instruments to get the values it needs. But which instruments?
How does (or how can) the location object find the instruments it needs
to instantiate itself?
The airplane provides the context for the request, so we want to
interrogate the instruments associated with (that are a part of) the airplane;
therefore, we ask the plane for those instruments. The location could ask
the plane for each instrument in turn, but our code will quickly reveal this
to be a less than optimal design. For one thing, the plane is busy, and we
would be an annoying disruption. For another, we would have to add a
lot of responsibilities to the airplane of the sort, provide access to altimeter.
Instead of creating a provide access to responsibility for each instrument,

(continued)
 Chapter 7 Discovery 211

Another Example (continued)

we introduce yet another object—an instrumentCluster—and add a provide

access to instrumentCluster responsibility to the airplane.
An instrumentCluster is basically a collection of instruments. As with
any collection, we can ask it for a subset of its members and ask it to iterate
across its membership. The location can then ask the instrumentCluster for
the next instrument in the collection and ask that instrument for its value.
Remember, the thinking and design activities that we just discussed
take place as part of the activity of writing tests and code. It need not be
an explicit separate step in the development process. It’s reflective of the
thinking that goes on as each pair interprets their story, identifies the
objects needed, thinks about which object will do what, and how that
object will actually do what it is expected to do.
If the development team chose to actually draw models to assist them in
this chain of thought, they might look at Figure 7-11 (side 1 of the object cubes
for Airplane, aLocation, instrumentCluster, and Instrument) and Figure 7-12
(an interaction diagram confirming the report current location story).

Figure 7-11 Side 1 for airplane, location, instrumentCluster, and

F07MQ11

instrument objects.

(continued)
212 Object Thinking

Another Example (continued)

Figure 7-12 Object communications required for an airplane to

F07MQ12

tell the control tower where it is currently located.

Heuristics
Implicit but perhaps not obvious in the preceding section (and in the sidebar)
are a number of heuristics for discovering and assigning object responsibilities.
There are also some points where the developers’ thought processes might
have proceeded in a different direction: there are different and equally valid
 Chapter 7 Discovery 213

object designs. Some hints of why the examples shown are as they are can be
incorporated into the discussion of the heuristics.

Heuristic: Let objects assume responsibility for tasks that are wholly or
completely delegated to other objects in cases in which the responsibility
reflects natural communication patterns in the domain.
For example, the question might arise as to why the control tower asks the
airplane for a location instead of talking directly to a location object and hav-
ing it talk to the airplane’s instrument cluster. After all, it is the location that
actually does the work. If you are familiar with modern aviation, you know
that control towers have the ability to talk to at least one instrument carried by
an airplane—the transponder. This question raises a general issue of when an
object should or can assume responsibility for work actually done by others.
In the case of Airplane, it seemed a natural call. After all, the control tower
already has to talk to the planes (actually the pilots) about all manners of busi-
ness, so it doesn’t seem illogical for it to do so for location as well. The plane
is already “visible” to the control tower, and the plane provides the context of
meaning for the values that will be accumulated by the location object. It does
no harm to let the plane front for location, and it does a lot of good. Similarly,
there is value in letting the location object front for the instrumentCluster and
Instrument objects. If at some point in the future we have a more abstract
FlyingThing, we can use the same interface to obtain locations for any special-
izations of that class, such as the airplane. If the definition of location (the type
and number of values it contains) varies, the interface remains the same. If new
types of instruments are invented, they can be added to the instrumentCluster
without any need to know details about them—as long as they can respond to
the currentValue message.

Heuristic: Delegate responsibilities to get a better distribution and

increase reusability.
An example of this heuristic arises from the answer to the question of why the
Airplane needs the instrument cluster. We can arrive at the answer by enumer-
ating all the responsibilities that are involved with keeping track of and using a
set of instruments:
■ Know which instruments are included in the set.
■ Assemble a location by asking the correct instruments, in the correct
order for their values.
■ Remove an instrument from the set if it is bad.
■ Keep track of which instruments belong in which kind of airplane.
■ Add new instruments when available.
214 Object Thinking

■ Monitor the status of all the instruments.

■ Replace an instrument with an improved model when asked.
■ Make one or more instruments available when asked.
This seems like a lot of work for an airplane to undertake, especially
when we would like the airplane to concentrate on moving safely from one
place to another. Safe flying requires a lot of responsibilities in its own right,
responsibilities quite dissimilar to those just enumerated. It will also increase
reusability of both the airplane and instrument cluster objects by keeping sep-
arable responsibilities separate.
When an object accepts responsibility for work that it cannot do alone, it
does so to meet expectations expressed in its environment. As the programmers
write the tests and code for that kind of responsibility, each segment of code
represents their thinking about specific tasks that the object must complete if it
is to accomplish what is expected. Each of those tasks—code segments—might
be moved (via refactoring) to other objects better prepared or better suited to
fulfilling that specific task. Delegation arises from refactoring code. But it’s
expected that programmers learn from experience and that the next time you
think about a test or a block of code, you might recognize immediately that it
belongs in a different object—a kind of anticipatory refactoring that is grounded
in your own programming experience.
Delegation can lead to the temptation of management. This is to be
avoided. If you delegate, delegate. Do not try to retain any control over how your
delegate does its job. Do not attempt to second-guess your delegate’s work by
adding all kinds of elaborate result-checking code. When you actually code the
delegate object, give it the ability to evaluate its own work and correct its own
mistakes, and then rely on it to do its job while you concentrate on yours.

Heuristic: Use anthropomorphization and foreshadowing to determine

whether an object should assume a given responsibility.
Experience doing object development allows you to anticipate (or at least think
about) what might happen as a guide for current activities. Foreshadowing is a
necessary part of iterative development. It allows you to learn from prior experi-
ence and apply that learning to avoid doing “wasteful” or “inappropriate” things
right now. The future you anticipate might be the next test, the next line of code,
the next object or task in the story you are working on, or even the next story.
Foreshadowing will frequently take the form, “If I give this object this
responsibility, later I will find that it requires this and this piece of information
and will require an ability to do this thing in this manner.” You can then decide
that the responsibility should or should not, in whole or part, be assigned to the
object under consideration. If your immediate thinking about a method seems
to be leading you in the direction of constructing a complicated and long
 Chapter 7 Discovery 215

method, stop! Think ahead a bit and reevaluate what you are doing. There is no
need to write ugly methods if a bit of foreshadowing can help you refactor as
you write. Decisions regarding responsibility assignment and method code are
seldom final; they are just a way to help you distribute responsibilities in a way
that minimizes, to some degree, future work.
Foreshadowing is not a substitute for refactoring and emergence; rather, it’s
a way to learn from experience and to apply XP and object thinking principles as
you work. For example, as you start writing the twentieth line of code for the
method you are currently working on, you should be hearing alarms—“This is
too hard”—that cause you to pause and see whether you can mentally imagine
what the entire method will look like and refactor the image of the code instead
of the actual code.
Anthropomorphization assists us in thinking about responsibilities and other
objects in several ways. We can think of objects as being “lazy” and wanting to do
only a limited and closely related set of activities. We find the responsibilities that
need to be delegated to other objects by foreshadowing what might be involved
in fulfilling a responsibility. We can think about how much the object might need
to know, whether it will need to know the “wrong” kinds of things about other
objects (that is, anything not evident in the object’s interface), and what it might
need to do.

Heuristic: Responsibilities should be distributed among the community

of objects in a balanced manner.
No object should have a disproportionate share of the responsibilities. How is
it determined whether an object has too many or too few responsibilities? If you
use a 3 × 5 lined index card to record your responsibilities—one to a line—you
will find you are limited to six or seven entries. This is an excellent heuristic for
the maximum number of responsibilities that an object can or should assume.
The lower bound is that an object must have at least one unique responsibility,
or there is no justification for making it a separate kind of object.
A balanced distribution of responsibilities across a group of objects is the
goal. Also implied is a balance in the distribution of knowledge among those
objects. An object should not do too much or know too much. You can determine
the knowledge distribution, roughly, via foreshadowing, as previously noted.

Heuristic: Always state responsibilities in an active voice describing a

service to be performed.
A common tendency for new object developers is to list as responsibilities
things that an object must “know.” An airplane must “know its current location.”
While this may be true, it can be misleading. It implies too much about how the
object might be implemented because “knows” implies an instance variable. It’s
also quite possible that an object will know things (a private key for decryption,
216 Object Thinking

perhaps) that it will not be willing to share with others and that therefore will
not be included in the interface for that object. Provide private key would not
appear as a service, although the message privateKey might be in the object’s
protocol with the designation that it is a private message. Decrypt message, on
the other hand, might be a listed responsibility. Always state your responsibili-
ties in terms of a service, with an awareness of a possible client for that service.

Heuristic: Avoid responsibilities that are characteristic specific, that

focus on providing a potential user with the value of a single
characteristic of the object.
It would be a mistake to list responsibilities such as know age, know height,
know eye color, and know weight. (See the preceding heuristic.) It’s also a mis-
take to list report age, report height, report eye color, report weight. Even though
the latter are in an active voice, they are characteristic specific and should be
avoided. The source of the problem is that an object might have a lot of charac-
teristics, and you have room for only six responsibilities. It’s better to look for an
umbrella label for the responsibility, such as describe yourself or identify yourself.
Important advantages will be gained by following this heuristic, that is, keeping
the conversation at the right level: what, not how or how much.

Note As a developer, you try to get basic behavior first. Obviously,

when you get to the details, the actual details will be necessary, but not
now. For example, you might have a responsibility called describe your-
self or provide access to demographic information. You might ask the
on-site customer a couple of questions:
“When you say ‘demographic information,’ do you mean name,
address, phone number, and so on?” “Yes,” she might reply, “that, and
hair color, eye color, and the like.”
“Are they all used just for informational purposes, or are some of
them prominent in other behavior?”
“They are mostly informational, except … ,” an answer that may
lead to a new responsibility or a natural grouping.
Sometimes you get domain people who want you to know every
characteristic. Assure them that “you’ll get the complete list when we
are getting ready to implement this behavior.”

One advantage is flexibility. If you make the interface to your class depen-
dent on the characteristics it might need to report and someone decides on a new
characteristic, you will need to redefine your class. The alternative is to have a
 Chapter 7 Discovery 217

responsibility describe yourself, which results in giving the requester a description

object. This description object might be a simple collection of key/value pairs or
an arbitrarily complex object with its own domain-driven (not just this story) set of
behaviors. Even the simplest case provides significant flexibility. The description
object can be asked for one or more of the values it contains so that the client gets
the information it needs. However, when you decide to add a new characteristic to
an object, you can do so by simply sending a message to its description object
telling it to add a key/value pair reflecting the new trait. (Again, assuming a very
simple form of description object.) No redefinition or recompilation is required.
Foreshadowing implementation for a moment reveals another advantage
arising from the fact that different clients might need to ask about different
characteristics. If the characteristics are defined in the interface, a situation
requiring multiple object types (classes) is created. For example, it’s likely that
a credit department will need to know characteristics of customers that are
not shared by other departments in a company. The temptation, when charac-
teristics are part of the interface (or when a data-driven approach to design is
used), is to create a small class hierarchy with Customer, CashCustomer, and
CreditCustomer, with Customer defining common characteristics as attributes
and CashCustomer/CreditCustomer adding at least one unique attribute each.
If the characteristics of a customer are held in a description object, only
the Customer class is needed. All clients can ask the customer for its description
and then ask the description for those characteristic values in which it’s inter-
ested. Cash or credit becomes just an additional characteristic of the object, not
a determinant of class structure or other implementation details.
There are some obvious exceptions to this heuristic. Almost all objects will
need to identify themselves on occasion. The object returned when the “ID
please” message is sent might vary significantly from object to object: the object
might be as simple as a string or as complex as a passport. Depending on the
context, both types of identification object might be required. It’s quite likely
that if a simple identity string is a characteristic of the object, that same string
will likely be part of the more complex object, the passport. Deciding whether
a characteristic merits a separate responsibility will be revealed in the collection
of scenarios in which the object is involved. If an object has 10 characteristics,
for instance, and 1 of them is asked about frequently and by a number of
clients, it should probably be a separate responsibility. If the other 9 are the
subject of only occasional inquiries or are almost always asked for at the same
time and in the same circumstances, they should be grouped in some fashion,
and the responsibility becomes to return the group.
218 Object Thinking

Heuristic: Create proxies for objects outside your domain that are
sources of information required by objects within your domain.
In the process of completing both the binary use cases and the scenarios, a
number of objects that are outside your domain will be identified. Legacy sys-
tems and human beings are the most obvious examples. They continue to be of
interest and importance because they are sources of information or behavior that
your objects will require. These requirements will be exposed in the scenarios as
messages sent to these outside objects.

Note Some objects that appear frequently in your stories may turn
out, in fact, to be external objects or representatives of external
objects. For example, the airplane object in the air traffic control example
earlier in this chapter is really a kind of proxy for objects that are not
directly part of the ATC system itself. Real airplanes (their pilots,
really) are customers of the ATC and a source of information that the
ATC uses in fulfilling its responsibilities.

For instance, if I were an order object I might need to obtain information

and decisions from a human being. I will want them to describe themselves to me
and tell me where they live, and what items from our inventory they want. Since
you cannot build a human being object, you can create a HumanInformation-
Interface object or a HumanProxy object that will be your interface to the real-
world human. A generalization of this heuristic is a class DataSourceInterface, of
which HumanProxy is an instance. So common is the need for this heuristic that
there is actually a pattern—Proxy—that documents its structure and use.

Heuristic: Look for components.

A component will look exactly like an object at this stage of development; that
is, it will have a name and a set of responsibilities. When you foreshadow those
responsibilities, however, the object starts to look complicated and will seem to
need a lot of collaborators, usually a sign of a need for further decomposition
and redistribution of factored responsibilities.
If the set of responsibilities are strongly coherent and if domain experts
seem aware of but unconcerned by an internal complexity—the way people
are aware that an automobile has a complex internal nature but ignore that in
favor of using a car as a car—you probably have a potential component. You
should identify your object as a component and plan to do detailed analysis of
it later, as if it were a small, separate piece of application software.
Thinking Toward Design
Discovery involves applying object thinking to the problem of decomposition
(finding the objects that reside in the domain) and the problem of requirements
(the behaviors expected of individual objects and aggregations of objects as
they interact in that domain). Tangible results of discovery probably include
partially completed object cubes and a set of stories. These results take tangible
form because they have value as a kind of external memory for the group of
individuals involved in development.
The CRC card method of Beck and Cunningham moved directly from dis-
covery into implementation. Once you had a set of cards and stories, you
started to write code, doing some design-oriented thinking about the code as
you went along. XP follows the CRC card method closely: pick up a prioritized
story card, grab a partner, and start writing code. XP does require you to write
code to be used in testing your work before actually writing the code to imple-
ment your objects and stories. Design, in XP, is not a separate process; it’s
infused in the process of writing tests and code. (Actually, discovery is infused
in the same activity: you identify your objects as you begin thinking about
implementation—tests and code—of your user story.)
Moving directly to the coding process may work for those whose brains
are already saturated with object thinking, who have internalized the ideas and
ideals to the point that object thinking is automatic and largely nonconscious.
For those just learning object thinking (or XP), there is a potential pitfall: the
metaphors and definitions that served you well in discovery cease to provide
valuable guidance as you think about design and implementation. Even worse,
the “good” metaphors are replaced with “bad” metaphors, metaphors reflective
of computer thinking as discussed in Chapter 1, “Object Thinking.” One of the
roles of an XP coach is to provide direction and help novice developers learn
the skills of object and XP thinking.

219
220 Object Thinking

Thinking toward design suggests the continual application of object ideas

and metaphors as the developer’s focus shifts from story to specific object, to
specific method, to test, to code. Attention shifts from what to how, and how
involves making decisions about design and implementation that are not
required when the focus is on the domain and what is expected from objects
participating in that domain.
Thinking toward design implies that it isn’t necessary to actually construct
formal, documented designs—only that your actual activities should continue to
be guided by object ideas and metaphors. Tangible documents and models sug-
gested in the rest of this chapter are intended to illustrate object thinking about
design and implementation and provide a convenient means of sharing ideas
among a group of developers. As you internalize object thinking, the use of
tangible documentation becomes less critical while remaining useful as a kind
of checklist you can use as a substitute for perfect memory (both individual and
collective). An XP test is another form—the preferred form, usually—of external
memory and the only form that’s intended to persist past the point of its immediate
creation and use. Tests remain even after the whiteboard full of diagrams is wiped
clear and the deck of story cards has found its way to the recycling bin.

Object Internals
Discovery has provided a decomposition of a domain into objects, an assign-
ment of responsibilities to those objects, and a description of how objects might
communicate and cooperate with one another to complete tasks beyond the
capabilities of individual objects. This is usually not sufficient information to
actually build computer software simulations of those objects and of those
interactions.
Here are some examples of why this is so:

■ Sides 1 and 2 of the object cube (and the classical CRC card) capture
only behavioral expectations of your objects—they tell you nothing
about the internal construction of those objects and nothing about
the means used by those objects to fulfill their responsibilities.
■ Assignment of responsibilities to an object tells you nothing about
how to invoke that responsibility and nothing about what form the
response will take.
■ Identifying a need for collaboration tells you nothing about how the
client object uses its collaborator, nor anything about how it knows
of the collaborator’s existence and location, nor anything about the
form of communication with that collaborator.
 Chapter 8 Thinking Toward Design 221

There is a clear need to think about how your objects will be simulated, to
make decisions about construction (design). Those decisions must be guided
by the principles of object thinking and based on the intrinsic needs of the
object. An intrinsic need is defined as the means for fulfilling an assigned
responsibility. It’s necessary to discern additional information about our objects
and their needs. It’s inevitable that the process of determining this information
will involve making decisions about implementation, normally an activity asso-
ciated with design.
There are innumerable ways to implement even the simplest object or
object responsibility. Take, for example, obtaining the string that uniquely
identifies an object. Different languages provide different implementations
(object.id, the dot notation of Java and Visual Basic versus the id getter message
convention of Smalltalk); different groups set different naming conventions,
which necessitate different syntax; and even the nature of the “id” itself might
vary, in some cases being an object, in others a type.
While necessary, design decisions can impose unwanted limitations on
the ability to use an object in different contexts. This is reflected in the immense
difficulty involved in creating libraries of reusable components and the need for
extensive middleware anytime objects created in one context need to be used
elsewhere.
Keeping the object’s domain-assigned responsibilities in the forefront of
your thinking while making design decisions ameliorates design-based limitations.
For example, if you are designing an object for use in a database environment,
you might be tempted to implement the object so that its internal structure
mirrors the table structure of the database. This will limit the use of such an
object to that specific implementation environment—unless you employ middle-
ware or other kinds of interpreters and translators.
If, on the other hand, you continue to let your design be guided by the
domain and the behaviors allocated to an object by that domain, you can avoid
making decisions that will inhibit the natural composability of that object as it
exists in the domain. The following discussion attempts to illustrate the impor-
tance of domain-driven design while providing some guidelines for how to
accomplish it.

Knowledge Required
If we think of an object as if it were a human being (anthropomorphism) and
we give that object a task, it’s appropriate to think about what that object may
need to know to complete its assignment. We can use our understanding of
human beings and what they need to know to perform tasks as a metaphorical
guide for our thinking about object knowledge.
222 Object Thinking

For example, suppose we ask Sara to ride her bike to the store and bring
home some milk. She will need to know

■ How to ride a bicycle.

■ Where the store is located.
■ A route between home and the store.
■ Perhaps, if more than one store is nearby, which store we want her
to go to.
■ The actual quantity of milk required—disambiguate the word some.
■ If timing is an issue, when we want her to perform this service.

Actually, the list of things required is potentially very large. A lot of the
items on our list are assumed to already exist—for example, how to ride a bicycle.
We have criteria (usually tacit) that limit our listing of what Sara needs to know
to a few details. We have similar criteria for considering what an object needs
to know to fulfill its responsibilities.
If we ask Sara to ride her bike, we probably already know she has claimed
that ability and we trust her in making that claim. If an object says it can identify
itself, we assume (trust) that the object has a method—a block of computer
code—that actually performs the identification service. We do not need to
explicitly list such abilities as required knowledge. (We will explicitly name the
methods themselves when we specify how an object’s responsibilities are to be
invoked, that is, when we specify a message protocol.)
We will usually assume that Sara knows which store to go to and the route
to take unless she has more than one option and it matters to us which option
she selects.
We will also assume that now means now and that the service is to be
performed immediately upon receipt of the request—again, unless we want
the option of requesting the service at a specified time or in a particular set of
circumstances.
Most of the time, we are interested in recording the information—
knowledge—required by the object in order to fulfill its advertised services.
Information is perilously close to being what we typically think of as data.
It’s therefore very easy to fall into the “computer thinking” trap of assuming
that knowledge required equals object data structure or object attributes.
Object thinking will help us avoid this trap.
The responsibilities recorded on side 1 of the object cube drive our thinking
about the knowledge required. Look at each responsibility, and ask what the
object will need to know to fulfill this task. List your answer—as a descriptive
noun or noun phrase—on side 4 of the object cube. Figure 8-1 shows sides 1
 Chapter 8 Thinking Toward Design 223

and 4 of the airplane object cube introduced in the “Another Example” sidebar
in Chapter 7, “Discovery.”

Airplane (1)

id self
describe self
report current location instrumentCluster
move to new location

Airplane (4)

id (v) a String
description (v) a ValueHolder
instrumentCluster (v) an Object
new location (A) a Location
current location (c) a Location

Figure 8-1 Sides 1 and 4 of the Airplane object cube, showing the
F08MQ01

relationship between responsibilities and knowledge required.

How did object thinking lead to the results recorded on side 4 of the
object cube for the airplane?

■ The airplane is responsible for identifying itself. It therefore needs to

know its ID. We record the noun (actually an abbreviation) id on
side 4. (See the “Object Cube Idiom” sidebar coming up for addi-
tional explanation of why things are recorded the way they are in the
examples.)
■ It needs to report its current location—hence currentLocation.
■ It needs to move to a new location on request—therefore, newLocation.
■ Because side 1 of the object cube indicates that serving a request
for current location requires collaboration with the airplane’s
instrument cluster, it needs to know of the instrument cluster
itself—so we record instrumentCluster as a piece of knowledge
required.

In most cases, the list of knowledge required will be fairly short and rea-
sonably obvious. In some cases, a single responsibility might yield more than
one piece of required knowledge.
224 Object Thinking

Completing a list of items to be known is but the first step in thinking

about the object’s knowledge requirements. We have two other decisions to
make about each piece of knowledge recorded in our list: how it is obtained
and what form it takes (what class will be used to encapsulate that knowledge).

Object Cube Idiom

A number of factors play a role in deciding the actual form—the actual words
and symbols—used to record information on an object cube. Paramount
among these is the need to be explicit and avoid ambiguity. For example, the
expectations of an object should be obvious from the phrases selected to
record those responsibilities; the name of a piece of knowledge should
unambiguously describe the semantic understanding of that knowledge; the
names given to classes and methods should reflect the essence of those classes
and methods. (In this regard, we are very Confucian in our insistence that
“only if things are given the proper names will all be right under Heaven.”)
Countering the need for explicitness is the need for brevity. Eventu-
ally most of the names will be used in writing program source code. No
coder really likes to type long descriptive names.
Another factor is the syntax of the programming language that the
development team is most familiar and comfortable with. Smalltalkers will
be quite comfortable recording id: aString as a method name or using
Integer as a class name, but C++ programmers are more likely to use
id(string) and int in similar circumstances.
This author, like everyone else, is a victim of his past: the idiom I am
most comfortable with derives from the use of the Smalltalk programming
language and its associated style and idiom. Although I will try to be as
non-language-specific as possible in my illustrations, be forewarned that
some idioms and conventions will inevitably creep in.
Another example of idiom is naming conventions such as the one
that suggests that the method names for retrieving the object in a named
variable and for placing an object in that variable are the same as the vari-
able name itself—with the addition of an argument in the case of the put
method. For example, the airplane has a piece of knowledge named id.
A method for retrieving the object encapsulating that ID (usually a string)
would be simply id. The method name for replacing the id string with
another string would be id(aString).
The proper idiom for use on object cubes, in code, and in any other
phase of modeling or development should reflect the community doing

(continued)
 Chapter 8 Thinking Toward Design 225

Object Cube Idiom (continued)

the development. It’s the responsibility of the developers in your organi-

zation or your domain to determine appropriate idiom and to train new
members of your development community in the use of that idiom. XP
practices of coding standards, pair programming, and collective owner-
ship of code support the creation and dissemination of appropriate idiom.

An object has any of four different ways to gain access to the knowledge
it requires:

■ It can store that knowledge in an instance variable.

■ It can ask (in the specification of a message signature) for that knowl-
edge to be provided along with the request for service (the message).
■ It can obtain the knowledge from a third party—another object.
■ It can “manufacture” the knowledge at the point when it is required.

Upon deciding which of these options is most appropriate, we will record

that decision by noting an appropriate symbol next to each piece of knowl-
edge. The symbols used are arbitrary, but in the examples in this book we will
use V for variable, A for argument, C for collaboration, and M for method. Here
are several of heuristics for deciding which option to use:

■ Objects are lazy. Every time you decide to store a piece of required
information in a variable, the object must assume responsibility for main-
taining that variable; it must add the capability to retrieve and to update
the contents of that variable upon request. So whenever possible and
appropriate, use argument (A) and method (M) instead of variable.
■ Collaboration is a form of dependence. Deciding to use a collaborator
to obtain required knowledge makes you dependent on that collab-
orator. Objects strive for independence, so collaboration should also
be minimized to the greatest extent possible.
■ Remember the definition of collaboration—requiring the service of
an object not found inside your own encapsulation barrier, which
will eventually require direct or indirect coupling with that collabo-
rator object. (With direct coupling, you know the actual name or ID
of the object used as a collaborator, whereas with indirect coupling,
you know where to find the object—in a global variable, perhaps.)
Coupling of this sort is just as undesirable in object thinking as in any
other development method or approach.
226 Object Thinking

■ In addition to knowing who your collaborator is (for example, the

instrumentCluster piece of knowledge recorded on side 4 of the Air-
plane object cube example), you will also need to know the message
that will be sent to the collaborator in order to invoke its help. That mes-
sage might require arguments and, if so, those will have to be added to
your knowledge required list. Using collaboration therefore increases
the intrinsic complexity of the object using collaboration as a means of
access as well as establishing an undesirable coupling between objects.
■ You might want to add the symbol G as an option for recording your
decision to access a piece of knowledge. G would stand for collaboration
with a global variable. Starting down this road, however, might lead to
making a distinction between a global in an application (Ga); a system
global—the system clock (Gs), for example; or some sort of intermediate,
such as the pool dictionaries in Smalltalk (Gp). If making such distinc-
tions helps you and your group, there’s no harm in using them. Take
care that your object cube does not reflect a particular implementation
context to the point that the general utility of the object is lost.
A final decision about knowledge required is that of determining the kind
of object that will embody (encapsulate) the information. A class name will be
recorded for each piece of information listed on side 4 of the object cube.
(Using a class name rather than a program language type is preferred for this
purpose because some information is complex—that is, not a simple primitive.)
There should be one encapsulating object per knowledge item.
In some cases, you will discover an entirely new kind of object when making
this decision—the location object, for example, used to encapsulate the various com-
ponents that make up an airplane’s location: altitude, latitude, longitude, and vector.
Discovery of the location object is a direct result of applying object thinking
to the question of knowledge required. As you think about what a location really
is, you discover the various component values that make up a location. You apply
the “lazy object” principle and find out that both the airplane and the instrument
cluster find that keeping track of the location components is too taxing and needs
to be delegated to someone else. The other candidate is an instrument, but no
instrument should know or try to keep track of any value except the one specifi-
cally generated by that instrument. Since none of your known objects is a suitable
home, you will likely decide to create a new object with responsibilities for
recording and maintaining the values that make up a location.
Figure 8-2 shows sides 1 and 4 for the objects in the air traffic control
(ATC) example from the “Another Example” sidebar in Chapter 7. Except for the
location object just discussed, the selection of an encapsulating object is pretty
much a matter of common sense coupled with knowledge of the existing or
planned class library for your domain.
 Chapter 8 Thinking Toward Design 227

Figure 8-2 Sides 1 and 4 of the objects (classes) introduced in the ATC
F08MQ02

example from the preceding chapter.

Figure 8-3 shows side 1 of the objects in the mortgage trust application
discussed in Chapter 7, and Figure 8-4 shows the knowledge required for those
objects.

Message Protocol
Side 1 of the object cube tells us what an object can do but reveals nothing
about how to ask for the advertised services. We know from object thinking
in general that services are invoked by the sending of a message to the object
providing the service, but what form must the message take? This is not a trivial
question because the form of the message is arbitrary, but it must be exact, or
the receiving object will ignore it. (Actually, it will cause an error if the message
is wrong—a variation of “I haven’t the foggiest notion what you are asking
me to do.”)
 228

F08MQ03
Class: Object Class: LIMT Class: Payment
Object Object Class: DataSourceInterface
display self Object
fund mortgages calculate self
id self invest surplus return calculation amounts return parsed value
persist return expense total indicate value change
grant access return surplus total implement collection behavior
Object Thinking

describe self return income total

dispatch state changes

Class: Form Class: Collection Class: Mortgage Class: Investment

Object Object Object Object Class: Employer
fill self in add element return application id return periodic income
EntryField delete element return couple return current value return weeks employee worked

in the preceding chapter.

validate self return subset of elements return home return base value return wages employee earned
return status return current gain/loss
indicate terms return object invested in

Class: MortgageApplication
Class: ValuedCollection Class: ValueHolder
Form, Object
Object Object Class: Bank Class: InsuranceAgent
return funding status
return cumulative value
return keys
implement collection behavior return savings amount provide insurance premium
return value at key
implement collection behavior return checking balance
return customer rating
return accounts for customer

Class: EntryField
Object Class: InvestmentCollection
Object, ValuedCollection Class: CalculationElement
Class: Home Class: Person
obtain value
validate self return periodic income
calculate self
return validation rules return sale price provide employer
return held class name return area provide bank
return legal description

Class: SelfEvaluatingRule
Object
evaluate self Class: Assessor Class: Couple
Payment Collection Mortgage Collection

return tax assessment return bank

return median price for given area provide marriage date
provide person

Figure 8-3 Object classes from the mortgage trust application introduced
Expense Collection
 F08MQ04
Knowledge Required: Object Knowledge Required: LIMT Knowledge Required: Payment
current display: aFrame Argument Knowledge Required: DataSourceInterface
expenses: aMoney Method calculation elements: aCollection Variable
id: aString Variable surplus: aMoney Method segmentation rule: aSelfEvaluatingRule Variable
calculation item: aString Argument
diskAddress: aDiskAddress Argument amountOnHand: aMoney Variable state: anEvent Variable
changedState: aBoolean Variable income: aMoney Method view: aView Variable
description: aValueHolder Variable myExpenses: aCollection Variable returnable elements: aCollection Variable
changed: anEvent Method my income sources: aCollection Variable
application: aMortgageApplication Argument
amount to be funded: aMoney Method

Knowledge Required: Form Knowledge Required: Collection Knowledge Required: Mortgage Knowledge Required: Investment
Knowledge Required: Employer
my entry fields: aCollection Variable added element: anObject Argument application: anApplication Variable periodic income: aMoney Collaboration
my validation rules: aCollection Variable deleted element: aString Argument couple: aCouple Variable current value: aMoney Collaboration employeeId: aString Argument
selected criteria: aCriteria Argument home: aHome Variable base value: aMoney Variable weeks worked: anInteger Variable
status: aCharacter Variable gain/loss: aMoney Method wages: aMoney Variable
terms: aCollection Variable investment id: aString Variable employees: aCollection Variable
my exchange: anInvestment Variable

Knowledge Required: MortgageApplication

Knowledge Required: ValuedCollection Knowledge Required: ValueHolder
funding status: aCharacter Variable Knowledge Required: Bank Knowledge Required: InsuranceAgent
cumulative value: aMoney Method
key-value pairs: aCollection Variable savings amount: aMoney Method
my elements: aCollection Variable home legal id: aString Argument
customer rating: aString Variable premium amount: aMoney Method
checking amount: aMoney Method home description: aValueHolder Argument
accounts: aCollection Variable
customer id: aString Argument

Knowledge Required: EntryField

Knowledge Required: InvestmentCollection
value source: aDataSourceInterface Variable Knowledge Required: CalculationElement Knowledge Required: Home Knowledge Required: Person
value: anObject Collaboration periodicIncome: aMoney Method
value: aMoney Method employer: aString Variable
validationRules: aCollection Variable salesPrice: aMoney Variable
calculation formula: aSelfEvaluatingRule Variable bank: aString Variable
held class: aSymbol Variable location: aString Variable
Chapter 8

legal id: aString Variable

Knowledge Required: SelfEvaluatingRule

Knowledge Required: PaymentCollection
Knowledge Required: MortgageCollection Knowledge Required: Assessor Knowledge Required: Couple
grantAmount: aMoney Method
home legal id: aString Argument bank: aString Variable

for the mortgage trust application introduced in the preceding chapter.

assessment formula: aSelfEvaluatingRule Variable marriage date: aString Variable
area: aString Argument spouses: aCollection Variable
median price table: aCollection Variable

Figure 8-4 Knowledge required for objects and responsibilities identified

Knowledge Required: ExpenseCollection
Thinking Toward Design
229
230 Object Thinking

We also have no clue about the way the object will respond to any request
sent its way. Will it provide us something in return? In many cases, we hope so.
If it does, what will be the nature of the returned item?
To answer these questions, we use side 5 of the object cube to record a
message protocol—a list of messages and their associated responses. As with
knowledge required, we refer to side 1 to elicit the necessary list of messages.
We also apply whichever idiom and convention for message syntax that’s
employed in our development environment. As individual messages are
recorded, care must be taken to maintain consistency with decisions made else-
where on the object cube—notably the names and encapsulating objects noted
on side 4 (knowledge required).
Figure 8-5 shows side 1 (responsibilities) and side 5 (draft message protocol)
for the objects in the ATC example. Figure 8-6 and Figure 8-7 show, respectively,
side 1 and side 5 for the objects in the mortgage trust example.

Draft Message Protocol: Airplane

id: aString
instrClstr
instrClstr: aCollection
currentLocation
moveTo: aLocation
description
description: aValueHolder
id

Draft Message Protocol: InstrumentCluster

add: anInstrument
delete: anInstrumentId
returnInstruments: selectCriteria
assembleLocation

Draft Message Protocol: Instrument

value
status

Figure 8-5 Sides 1 and 5 of the objects in the ATC example introduced
F08MQ05

in the preceding chapter.

 F08MQ06
Class: Object Class: LIMT Class: Payment
Object Object Class: DataSourceInterface
display self Object
fund mortgages calculate self
id self invest surplus return calculation amounts return parsed value
persist return expense total indicate value change
grant access return surplus total implement collection behavior
describe self return income total
dispatch state changes

Class: Form Class: Collection Class: Mortgage Class: Investment

Object Object Object Object Class: Employer

(object cube side 1).

fill self in add element return application id return periodic income
EntryField delete element return couple return current value return weeks employee worked
validate self return subset of elements return home return base value return wages employee earned
return status return current gain/loss
indicate terms return object invested in

Class: MortgageApplication
Class: ValuedCollection Class: ValueHolder
Form, Object
Object Object Class: Bank Class: InsuranceAgent
return funding status
return cumulative value return keys
implement collection behavior return value at key return savings amount provide insurance premium
implement collection behavior return checking balance
return customer rating
return accounts for customer

Class: EntryField
Object Class: InvestmentCollection
Object, ValuedCollection Class: CalculationElement
Class: Home Class: Person
obtain value
validate self return periodic income
calculate self
return validation rules return sale price provide employer
Chapter 8

return held class name return area provide bank

return legal description

Class: SelfEvaluatingRule
Object
evaluate self Class: Assessor Class: Couple
Payment Collection Mortgage Collection

return tax assessment return bank

return median price for given area provide marriage date
provide person
Expense Collection

Figure 8-6 Responsibilities for the objects in the mortgage trust application
Thinking Toward Design
231
 232

F08MQ07
Draft Message Protocol: Object Draft Message Protocol: LIMT Draft Message Protocol: Payment
Draft Message Protocol: DataSourceInterface
id expenses calculate
id: aString amountOnHand amountAt: aString value
persistAt: aDiskAddress surplus segmentationRule: aSelfEvaluatingRule
busy income segmentationRule
displayOn: aFrame expenseCollection state
Object Thinking

restoreFrom: a DiskAddress expenseCollection: aCollection state: anEvent

busy: aBoolean investments view
registerFor: anEventwith: aRegistration amountOnHand: aMoney view: aView
elements
Draft Message Protocol: Form Draft Message Protocol: Collection Draft Message Protocol: Mortgage Draft Message Protocol: Investment
Draft Message Protocol: Employer
nextEntryField add: anObject application income

(object cube side 5).

entryFields delete: aString couple currentValue weeksWorkedBy: aString
entryFields: aCollection returnSelected: aCriteria home gainOrLoss weeksWorked: anIntegerBy: aString
validationRules nextElement status baseValue wagesFor: aString
validationRules: aCollection terms baseValue: aMoney wages: aMoneyFor: aString
instantiate application: anApplication exchange
couple: aCouple exchange: anInvestmentExchange
home: aHome investment

Draft Message Protocol: MortgageApplication

Draft Message Protocol: ValuedCollection Draft Message Protocol: ValueHolder
status Draft Message Protocol: Bank Draft Message Protocol: InsuranceAgent
cumulativeValue at: aKey
status: aCharacter
valuedElements savingsFor: aString premiumFor: aValueHolder
add: aKeywith: aValue
valuedElements: aCollection checkingBalanceFor: aString
delete: aKey
accountsOf: aString
keys
accounts: aCollectionFor: aString
at: aKeyput: aValue
rating
rating: aString

Draft Message Protocol: EntryField

Draft Message Protocol: InvestmentCollection
value Draft Message Protocol: CalculationElement Draft Message Protocol: Home Draft Message Protocol: Person
validate periodicIncome
formula salesPrice bank
heldClass
formula: aSelfEvaluatingRule salesPrice: aMoney employer
heldClass: aSymbol
calculate location bank: aString
validationRules
validationRules: aCollection location: aString employer: aString
legalId
legalId: aString

Draft Message Protocol: SelfEvaluatingRule

Draft Message Protocol: PaymentCollection
Draft Message Protocol: MortgageCollection Draft Message Protocol: Assessor Draft Message Protocol: Couple
grantAmount
taxAssessmentFor: aString bank
medianPriceOf: aString bank: aString
medianPriceTable marriageDate
medianPriceTable: aCollection marriageDate: aString
assessmentFormula spouses
assessmentFormula: aSER spouses: aCollection

Draft Message Protocol: ExpenseCollection

Figure 8-7 Message protocol for the objects in the mortgage trust application
 Chapter 8 Thinking Toward Design 233

Completing side 5 is usually straightforward. The following heuristics help

with any nuances involved:

■ A full message signature includes three elements: the message selec-

tor (the actual message name), arguments (if any; arguments are
optional), and the nature of the object returned to the sender of the
message. For example, includes (anObjectSpecification) aCollection.
Parsing this message signature results in identification of includes as
the message selector, (anObjectSpecification) as the argument, and
aCollection as the object returned. Because we are recording the
messages received by the object whose cube we are completing, we
omit the name of that object when drafting the message protocol;
only the last three elements of the message signature are recorded
on side 5.
■ Messages must be descriptive of the nature of the service (behavior)
being invoked. It should be possible for a naive user of your object to
immediately discern what is likely to happen if the indicated message is
sent to the object. This includes specification of arguments—it should be
obvious what kinds of objects are being passed as part of the message.
■ Messages must be relatively terse (programmers will get tired of typ-
ing long messages), a requirement that is at odds with and less
important than the descriptiveness requirement. Programs are read
for understanding more frequently than they are typed.
■ Messages should reflect the syntax of your chosen implementation
language and any standards or conventions, such as naming,
adopted by your development group.
■ If you have decided, on side 4, to store an object encapsulating
some bit of information in an object instance variable, you must
add two messages to your protocol: one to obtain the object in the
variable (a getter message) and one to replace it (a setter message).
Either or both of these messages can be categorized as private (see
the next section about message contracts), or you can decide to
make the variable immutable. It is also possible (relevant if either
the getter or setter is considered private) that your language allows
access to a variable by reference instead of via an explicit message.
(Languages that allow access by reference from outside the object
are suspect in terms of object thinking principles and philosophy.)
Nevertheless, it is a good idea to record both getter and setter on
side 5, even if you elect one of the other implementation options.
Two common conventions for getter/setter message style are to use
234 Object Thinking

the variable name itself as the getter and the variable name plus
argument as the setter; and to append the word get or set to the
variable name. (Example for the id variable: id and id: aString—or
getId and setId.) The argument, if any, will reflect whatever deci-
sion we made on side 4 as to the nature of the encapsulating object
for that variable.
■ Some messages are imperative commands: “Don’t bother giving me
anything back; just do this!” In those cases, no object is returned. A
Smalltalk convention is to note self as the object returned. For those
more familiar with C++ and Java, it is perfectly appropriate to put
the term void in place of self. The meaning is equivalent, although there
are still technical differences as to what, if anything, is actually returned.

Message Contracts
Side 3 of the object cube has not been forgotten—it’s only now that it will
make some sense to talk about what’s recorded on that face of the object cube.
That side 3 is not swapped with side 5 is a historical artifact: the idea of con-
tracts precedes the idea of object cubes. A contract is a concept introduced by
Rebecca Wirfs-Brock and her coauthors as an extension of the information
recorded on the original CRC cards invented by Beck and Cunningham. The
idea was to aggregate responsibilities—later, messages—into groups to reflect
the users of those methods. This particular use of contracts did not gain wide
acceptance, and the idea of contracts became rather obscure. Even so, the pro-
gramming concept of interfaces and templates has much in common with the
idea of contracts.
A concept from programming—message scope or visibility—provided
renewed use for contracts. In a programming language such as Java, methods
(and their invoking messages) could be designated public (anyone can send that
message and invoke that service), private (only the object itself can send the mes-
sage to itself), and protected (only a designated group of user objects can send
the message). Other languages, most notably Smalltalk, did not make explicit
provision for such method/message declarations. Using contracts to at least specify
the intent of the object developer as to the proper use of messages was a natural
extension of the notion of contracts. If the implementation language supported
message scoping, side 3 provided a specification to the programmer. If not,
side 3 documented the intended use of the categorized messages—and no good
object programmer would misuse private or protected messages. (Smile.)
As class libraries grew in size and the messages associated with individual
classes grew in number (something that should be minimized if object thinking
guides the design of those classes), it proved useful to create subcategories
 Chapter 8 Thinking Toward Design 235

of classes and of methods to simplify the search for an appropriate class or

method in the library. A typical browser in an integrated development environ-
ment (IDE) might show the following four lists: categories of classes, class
list, categories of methods, and method list. A user would select the desired
class category and see only those classes in that category displayed in the sec-
ond list. Then the user would select a method category and see only those
methods included in that category displayed in the last list. Typical method cate-
gories include accessing, displaying, updating, and calculating. Such categories
can be captured on the object cube as contracts on side 3.
The layout of side 3 (shown in Figure 8-8) is simple: a contract name followed
by an indented list of the messages intended to be included in that contract.
A message can appear in more than one contract unless the contracts reflect
programming specification of public, protected, and private.

Figure 8-8 Contracts for all classes in both the ATC and mortgage trust
F08MQ08

application examples (object cube side 3).

236 Object Thinking

State Change Notification

“Objects encapsulate state.” When this claim is made, it’s usually a reference to
changes in the objects occupying an instance variable. This idea reflects the
common definition of state—a change in value of any aspect or characteristic of
a thing—colored by the way state is used in data-driven object design.
If an object is properly designed, it is typically so simple as to have very
little interesting state. Some examples of state might include the following:

■ An object in an instance variable has been replaced with another

object; the object (or value of that object) matters far less than the
fact that the change occurred, so we would characterize the new
state as changed.
■ An object might be in the process of responding to a message and
therefore unavailable to receive a new message. State = busy.
■ An object might be defective in some sense, out of calibration or
completely inoperative, yielding the states faulty and dead. (In the
latter case, we are not looking for dead so much as “I’m dying, gasp,
gasp … the butler did it” written in blood with the object’s last exha-
lation. Or, less colorfully stated, failing.)
■ An object might be uninitialized: none of its instance variables contain
objects other than Nil.
■ An I/O device might have a ready state.
■ A state based on the value of an instance variable: the pressure variable
of a steamEngine object and the danger state that would occur if the
value of pressure exceeded some limit.
■ From the world of persistence (databases), a variation on the
changed state can be surmised: dirty, which means a change that has
yet to be reflected in the persistent persona of the object. There is an
inconsistency in value between the cell of a database table and the
object stored in an instance variable.

Although it is possible for an object to have numerous states, only a few

of those states are likely to be of any interest to anyone outside the object itself.
In those cases in which other objects might be interested in a state change, it is
appropriate to list and describe those states on side 6 of the object cube. The
syntax for side 6 is very simple—a descriptive name of the state and a short
description of that state. Figure 8-9 shows side 6 for all objects in the ATC and
mortgage trust examples.
 Events Generated: Object

F08MQ09
Events Generated: LIMT Events Generated: Payment
Events Generated: DataSourceInterface
busy bankrupt past due
changed insufficient amount unavailable

Events Generated: Form Events Generated: Collection Events Generated: Mortgage Events Generated: Investment
Events Generated: Employer
completed valueChanged

Events Generated: MortgageApplication

Events Generated: ValuedCollection Events Generated: ValueHolder
Events Generated: Bank Events Generated: InsuranceAgent
value updated

"changed" state inherited from class Object).

Events Generated: EntryField
Events Generated: InvestmentCollection
Events Generated: CalculationElement
Events Generated: Home Events Generated: Person
completed
divideByZero
Chapter 8

Events Generated: SelfEvaluatingRule

Events Generated: PaymentCollection
Events Generated: MortgageCollection Events Generated: Assessor Events Generated: Couple

Events Generated: ExpenseCollection

Events Generated: Airplane Events Generated: InstrumentCluster Events Generated: Instrument

Thinking Toward Design

hijacked lost power uncalibrated

emergency failed
lost communication

side 6). Note that most objects have no events. In some cases this is because they are
Figure 8-9 Events for objects in the ATC and mortgage trust examples (object cube

inherited (MortgageApplication from Form), but usually it just reflects that most objects
are not willing to share much state information, except for the almost universal
237
238 Object Thinking

Here are some important caveats concerning object state and side 6 of the
object cube:

■ Side 6 records only those states that the object is willing to make
visible to other objects. In this sense, side 6 is akin to side 1 in that it
publishes part of the object’s interface: the part of the object visible to
others. Some state changes might be kept private to better reflect the
domain being simulated by the object. For example, a selfCalibrating-
Instrument might have a state, out of calibration, that it does not
make visible to the outside world. Instead, it detects that state itself,
takes steps to correct that state, and only if it fails in such attempts
does it generate a public state, failed. Failed would appear on side 6
of the object cube, but out of calibration would not.
■ State changes are visible only to the object experiencing the change.
Many objects will choose not to allow anyone to be aware of their
changes, and even if an object does make a change public by list-
ing that state on side 6, it does not imply that you—or any other
object—can see that change, only that you—and all other objects—
can request to be notified when the object detects the change
in itself.
■ Side 6 of the object cube does not capture and is not intended
to capture state-related constraints on an object’s behavior. That
kind of information is captured in a static diagram—specifically a
state chart—and will be discussed in Chapter 9, “All the World’s
a Stage.”
■ Advertising a willingness to notify others of state changes implies
that the object has a mechanism for keeping track of who is to be
notified, how, and for what. At first this implied requirement might
seem to violate object-thinking precepts, but it’s quite possible to satisfy
such a requirement in a manner consistent with object thinking. The
mechanism is an eventDispatcher object. Every object capable (and
willing) of notifying others of its state changes must contain an event-
Dispatcher to effect that notification.

An eventDispatcher object can be visualized as a simple two-part table, as

shown in Figure 8-10. The first column of the table contains events, one per
row. The second column of the table contains a collection of eventRegistration
objects. An eventRegistration, in its simplest form, is a two-part object consisting
of a receiver and a message.
 Chapter 8 Thinking Toward Design 239

Event Dispatcher
Event Registration Queue
event 1 [REG1] [REG2]
event 2 [REG1]
event 3 [REG1] [REG2] [REG3]
event 4
event 5

Registration
[ receiverObject message ]
Registration (if priority allowed)
[ receiverObject message priority ]

Figure 8-10 An event dispatcher table and an event registration tuple.

F08MQ10

Creation of an eventDispatcher object increases flexibility by centraliz-

ing but not controlling the awareness of which events exist and who is to be
notified when they occur. If an object decides to publish a new event, a sim-
ple message to the eventDispatcher requesting the addition of a new row to
the table is sufficient to effect that change. An object’s eventDispatcher can
be queried as to which events are available, and it responds with a collection
(a list) of the contents of the first column. Such queries are almost always
made by the programmer as she decides what objects and scripts need to be
created to realize a story. Only if you were building a complex-adaptive sys-
tem (artificial life, for example) would you expect other objects to query a
dispatcher about its event list at run time.
If objectA wants to know about a state change in objectB (one that
objectB has advertised as public), it sends a registration of its own construc-
tion to objectB’s eventDispatcher. The objectA object decides what message it
wants sent to effect the event notification, which means that objectA can
change its mind—use different messages in different contexts—simply by ask-
ing that its earlier registration be replaced with a new one containing a new
message. This capability provides significant run-time flexibility, allowing
changes without necessitating any changes in code and subsequent recompi-
lation of objectB.
One other possibility to be noted: suppose we need to notify objects
in a particular order—for example, some objects need to be notified imme-
diately and the needs of others are less urgent. We can change (probably
subclass) eventRegistration to be a triple—receiverId, message, priority.
240 Object Thinking

We can then allow the collection that makes up the second column of the
eventDispatcher to be a sorted collection. As eventRegistration objects are
added, they are sorted according to their priority values. (Event dispatching
as described here is consistent with the observer/publish-and-subscribe
patterns published elsewhere.)

Object Appearance
It would be difficult to talk about objects without talking about the graphical
user interface (GUI). The Star project at Xerox PARC simultaneously advanced
local area networking, GUI design, Smalltalk (object-oriented programming),
and alternative input-output (I/O) modes (notably the mouse). Networks and
I/O were sufficiently esoteric and close to hardware design that they went
their separate ways, post-PARC. GUI design and objects, however, emerged so
tightly coupled that it’s often assumed that the primary (if not exclusive) use
of object design and object programming is the construction of graphical
interfaces.
Visual development environments (such as Microsoft Visual Basic) man-
aged to convey the impression that objects were the same as the GUI widgets
that appeared in the interface design toolbox. Once you had those widgets in
place on your form object, you did real programming—sans objects. (Yes, it’s
true that the objects are still there, but they are not emphasized or enforced,
with the net effect that most Visual Basic programs tend to be event-driven
procedural in nature and not OO.) Even Smalltalk IDE tools such as Parts (an
extension for Digitalk Smalltalk) and VisualAge (IBM) emphasized the utility
of objects for GUI building with far less emphasis on the objects behind
the interface.
A worse error was propagated when many of the early tutorials on object
programming introduced a misconception by suggesting a “method” colloquially
referred to as “cocktail napkin design.” It was suggested that the correct way to
design an application was from the interface in; that is, sketch your interface on
a cocktail napkin and then find the objects necessary to implement that inter-
face. The GUI became a straitjacket to which objects had to conform. This, in
effect, meant that object design and implementation were little more than the
hard-coded reflection of the specific and idiosyncratic design of a set of visual
interfaces. Change the visual interface, and you had to change the object. Not
only did this make for a lot more work, it meant that objects were not reusable
in different contexts if those contexts defined visual interfaces in an alternative
fashion.
 Chapter 8 Thinking Toward Design 241

Note “You can’t emphasize this point enough,” suggests Ken Auer,
speaking about the danger of “GUI-in” design. “In my early days of OO,
if you talked to someone who had been using Smalltalk for 1 to 3 years,
you could almost rely on the fact that they ‘got it.’ After the advent of
Parts, VisualAge, and (to a slightly lesser extent) VisualWorks, I had
many interviews with programmers who had been ‘programming with
Smalltalk’ for 2 or more years and didn’t realize there were supposed to
be some objects other than the GUI and the database. Even the ones
that recognized this tended to have a poor handle on how central the
domain objects should be because they probably spent a good two-
thirds or more of their time dealing with GUI and/or DB issues.” Ken’s
comments support a basic premise of this book: object thinking is most
easily corrupted by reverting to old, familiar forms of thought. Laying out
widgets on a form and objects as data entities are but two examples.

Object thinking acknowledges the special relationship between an object,

X, and some group of other objects whose role is to represent, visually or
otherwise, object X. At the same time, it suggests some important differences in
the way that relationship is discovered and implemented.

Occasions Requiring an Appearance

While engaged in the task of discovering objects and responsibilities, we told
stories about object interactions. Among the interactions in those scenarios
were instances of an object displaying itself to a client. An example might be a
character displaying itself on some medium for the benefit of an observing
human being. What actually appears on the medium is not the character object;
it is a representation of some aspect of that object (its visual appearance but not
its behavior, or dynamics, or “soul”), just as your photograph is not you but a
representation of some part (the surface part) of you.
The converse of displaying oneself to a client is to display the void at the
core of your essence when you have yet to be instantiated. This time you dis-
play the void as a kind of request for service, asking a human user to please
provide substance (a value) that will make you a fully instantiated object. An
entryField widget consisting of a simple white rectangle and perhaps a blinking
cursor within the bounds of that rectangle is an example of an object displaying
its “void” and requesting the human user to fill that void with a value. A hybrid
of both display and request is a display of your current value with the implicit
request to change that value if appropriate.
242 Object Thinking

Even the simplest object might have multiple representations of itself, just
as you probably have more than one photograph of yourself. A character, for
example, might have a boldfaced appearance and an italicized appearance. In
cases in which an object has multiple appearances and can be asked to switch
from one to the other, it’s necessary to add messages to the protocol that make
this ability evident. For example, a date object might have two messages in its
protocol—displayUS and displayEurope—the difference between the resulting
display being a transposition of the day and month values.
Most objects are compositions of simpler objects. Remember that every
instance variable an object might have contains another object. If an object with
instance variables has a gestalt (view of the whole) appearance, that gestalt is nec-
essarily a composite of the appearances of itself and all of its contained objects.
It’s easier to illustrate ideas about an object having multiple appearances
by using visual metaphors. But visual representations are not the only appear-
ances an object might have. Imagine an object that needs to store itself in a
relational database. The relational database management system (RDBMS)
cannot accept the object in its natural glory, so the object must marshal1 itself
into a stream of bits that can be accepted and held by the RDBMS. The resultant
stream of bits constitutes an appearance of the object but is not the object
itself, any more than a visual representation is.

Glyphs
The application of object thinking to the issue of text characters, numbers,
and graphical symbols should lead to the recognition that all of these are
really just instances of a single kind of thing: a glyph. A glyph object
would have one major responsibility: to display itself. To fulfill that
responsibility, it would need to know

■ Its origin, a point used in the definition of its extent and in

placement of its value on a medium.
■ Its extent, an area (not necessarily a regular area, but likely
some sort of polygon).
■ Its scaling factor, an analog of point size, some constant that
would allow it to occupy greater or lesser extent.

(continued)

1. Marshaling involves the object taking itself apart, asking each part to convert itself into bits, and then
asking each part to line up in some kind of order. The process is analogous to a modem converting
an analog signal to digital (and eventually back again).
 Chapter 8 Thinking Toward Design 243

Glyphs (continued)

■ Its orientation, a radian (used if the glyph is to be laid out other

than horizontally).
■ Its value, something as primitive as a bitmap or, more likely, an
algorithm (vector) that results in the generation of colored pixels
on a medium that is the actual appearance of the glyph.
■ Its ASCII value, a bit stream.
■ Its EBCDIC value, a bit stream.
■ Its Unicode value, a bit stream.

The message protocol for a glyph would include the display message
(the glyph would use values in all the appropriate variables to create
the bit stream sent to a printer or graphics card that actually generates the
display) and getter and setter messages for each of the instance variables
listed in the foregoing list.
If glyphs existed in typical information system applications, it would
not have been an issue when a certain rock star changed his name to a
Celtic-Egyptian symbol.

The separation of objects and views or appearances of objects is essential. This

separation of objects and views has also been addressed—but incompletely—in
a number of different ways in programming languages and architectural patterns
(for example, model-view-controller).
Object thinking would suggest that every object will have a collection of
appearances, will advertise (as a responsibility/service/message) its ability to
appear in different guises as well as variations (point size, for instance) of a single
guise. Each object will transfer to its appearance object (for example, a glyph,
as described in the sidebar “Glyphs”) responsibility for acquiring and maintain-
ing the information required to effect each appearance.
An elementary object, such as a character, might have a small set of
appearances each of which is a simple glyph. The appearance of a more com-
plicated object, such as a string, a date, or a number, would be a composite (an
ordered set of appearances of each of its constituent parts): each character in a
string, each number and symbol in a date, each integer and symbol in a number.
A still more complicated object, such as a form, would be a composite of its
own appearance (a boundary perhaps) plus the composites of each of the ele-
ments appearing on the form plus the composites and glyphs of each member
of each element that appears on the form.
244 Object Thinking

Keeping track of this apparent complexity is relatively straightforward:

you simply apply the composite pattern. Composite is one of the 26 patterns in
the first pattern book, Design Patterns. Applied to appearances, the composite
pattern asserts that every appearance is made up of glyphs or appearances.
A glyph is defined as a leaf node in the hierarchy; an appearance is defined as
an instance of appearance, which in turn is made up of appearances, glyphs, or
a combination. To display an appearance, you use recursion to traverse the
hierarchy, displaying all glyphs and decomposing all appearances into glyphs
and appearances until you reach a level containing nothing but glyphs.

Widgets, Forms, Reports

Visual programming languages typically provide a number of different
widget types to be used in collecting and displaying bits of information. A
character widget, a number widget, a date widget, and a currency widget
are examples. Object thinking suggests that there is a need for only one
type (class) of data entry widget. A dataEntryWidget would have an
appearance, the simplest form of which would be a simple rectangle or a
bit of underlining. It would have behavior that would include signaling
whenever it has been changed (a person entering a value in the box or
on the line). It would also possess a set of rule objects that it could use to
validate its own contents. An example of a rule might be valueEntered is a
date; if true, return True; else return False. Another example: valueEntered
is between minValue and maxValue. Because the rules are objects and
because each widget has a collection of rules that it can apply to itself—a
collection that can be modified at any time with add, delete, or replace
messages—creating any type of specialized widget is trivial. (See the dis-
cussion of self-evaluating rules in Chapter 10, “Supplemental Ideas.”)
A form, in light of object thinking, is nothing more than an ordered
collection of elements. An element might be a string or a dataEntryWidget.
A form would have responsibilities to display itself, collaborating with its
elements; update itself (add and delete elements); effect different
instances of itself (fill itself in, collaborating with its dataEntryWidget
objects); and validate itself, working with its own collection of rule
objects. Form validation rules exist for ensuring consistency among the
values stored in dataEntryWidget objects (each of those widgets having
already validated its contents according to its own set of rules). For example,
a form validation rule would ensure that the value stored in the Zip Code
widget is consistent with the value stored in the State widget.

(continued)
 Chapter 8 Thinking Toward Design 245

Widgets, Forms, Reports (continued)

A report is similar to a form in that each comprises a collection of

elements. The form adds and deletes elements to become different
instances of the form. The form displays itself by asking each element to
instantiate itself (obtain a value for itself) and display itself. Each element
has its own set of appearance rules, which it uses to display itself. An
example of an appearance rule is “My point size is 1.5 times the default
point size for this report.”
Both reports and forms make use of the glyph and composite patterns
discussed earlier in this section.

Object State, Object Constraints

Readers familiar with typical treatment of object development might be curious
as to the absence, so far, of any discussion of object state other than the will-
ingness to notify others of state changes as recorded on side 6 of the object
cube. Some definitions of objects suggest that they encapsulate state; and model-
ing methods and tools, such as Unified Modeling Language (UML), provide
sophisticated models for capturing state-of-the-object information.
Object thinking shows little concern about state when objects are dis-
cussed specifically because a properly designed object has very little interesting
state. What state it might have should be private—behind the encapsulation
barrier—except to the extent that the object is willing to make public the fact
that a state change occurred.
Most discussion about object state is really about state-based constraints to
be imposed on an object. Such constraints are not intrinsic to the object itself;
they are an aspect of the situation in which the object finds itself employed.
This kind of state modeling is important but not important in advancing our
understanding of individual objects. For this reason, discussion of state model-
ing involving objects will be taken up in Chapter 9.
All other constraints that might be imposed on an object are also reflective
of a situation, not of the object per se. Table manners, for example, reflect a set
of rules that promote or inhibit intrinsic human eating behavior—fingers and
communal bowl in an Ethiopian ethnic restaurant; a plethora of special-purpose
utensils and prescribed behaviors at a formal state dinner. It’s a mistake to
attempt to incorporate this kind of variation into an individual object’s specifi-
cation. This type of rule-based constraint will also be addressed in Chapter 9.
All the World’s a Stage
Chapters 7 and 8 focused on individual objects and problem domains. The dis-
cussion focused on discovering—identifying—the objects that already exist in a
domain and understanding what is expected of them in that domain. The focus
was on an entire domain—either an enterprise (Woodgrove Bank) or an area of
commerce (banking industry) as a whole—to avoid defining objects in terms of
their idiosyncratic behavior in narrowly defined applications.
Our goal of object discovery was facilitated by telling specific stories about
small groups of objects engaged in resolving specific problems or, collectively,
providing a specific service. We did this to illustrate, to explore, and to discover
the intrinsic capabilities and expectations of that object. No single story defined
the nature of an object. The set of stories we used to discover and model objects
was broad enough in scope to ensure a proper balance of specificity and gen-
erality: specificity so that the objects could be built and built simply; generality
to ensure that the same objects could be used in multiple contexts without
modifying their essential nature.
A number of concerns that typically arise in software development were
not mentioned or were mentioned but set aside for later discussion. Deferred
issues fall into four broad categories:

■ Static relationships, associations, affiliations, compositions, and inter-

action channels that are assumed to be, but seldom are, permanent
■ Scripting, providing the cues to be used by a collection of objects
while completing specific collective tasks

247
248 Object Thinking

■ Constraints, keeping objects from doing things that they are capable
of but that we want to prevent in certain circumstances
■ Implementation, providing detailed information about the inside of
our objects; algorithms to be used and detailed specifications about
formats and values of information

The first three categories were deferred because they do not inform us
about objects, only about contexts in which objects must operate. Implementa-
tion issues are deferred in accordance with the long-standing development rule,
“Figure out what must be done before you concern yourself with how it is
done.”
We are also being consistent with the Lego brick metaphor: separating the
problem of creating bricks from the problem of building things with bricks. To
reintroduce another metaphor, we separated our discussion of actors and their
intrinsic talent from the discussions of casting, screenplays, and direction. Our
goal was to create actors (objects) with the versatility required to assume many
different roles in a wide variety of genres and to avoid creating objects that
were typecast—doomed to play the same role, in the same screenplay, over
and over again.
Generally, we do not try to solve all the problems of the world in one fell
swoop. Instead, we arbitrarily set a boundary that delineates our focus and sep-
arates it from the rest of the universe. We call this bounded space our “system,”
or our “application,” or our “program,” or our “object.” Unfortunately, we also
tend to be a bit sloppy in our use of terminology and frequently use the terms
application, system, and program more or less interchangeably. In the rest of
this section, I will use the term application as a label for the domain and our
artificially bounded portion of that domain and the term application artifact for
specific instances of executing code—programs or objects. (To see the differ-
ence between application and application artifact, see the sidebar “Systems
and Artifacts.”)
 Chapter 9 All the World’s a Stage 249

Systems and Artifacts

Structured analysis was, in its own time, a new paradigm for software
development. As popularized by Larry Constantine, Ed Yourdon, and
Meillor Page-Jones, an initial task was the creation of a model of the
“existing system.” Developers were instructed to create a model—data
flow diagrams, entity relation diagrams, context diagrams, and even pro-
gram structure charts—to document their understanding of how things
were currently done. This effort was supposed to facilitate understanding
and modeling of a new “system” to replace the existing one. In practice,
this step was seldom completed: its value was not understood, and its cost
was significant. Later editions of books on structured analysis tended to
eliminate this step altogether.
Modeling the existing system has a parallel in object thinking:
domain-centric analysis. There is an implicit assertion in both: you cannot
begin to understand what must be until you understand what is. This
assertion has two corollaries:

■ Almost all of the objects you will ever need are already defined,
and already have behavioral expectations associated with them,
in the domain.
■ Almost all of the requirements of new development arise from
a misalignment of behaviors and objects. Misalignment results
when the wrong object (or group of objects) is providing a par-
ticular service, a service is more appropriately provided by a sil-
icon-based object simulation instead of a carbon-based
biological object, or, occasionally, no existing object is capable
of providing the needed service.

Christopher Alexander (Notes on the Synthesis of Form) offers

another expression of the need to understand the problem and the prob-
lem (domain) space in detail before proceeding with development.
Design, according to Alexander, is the process of adapting a solution to a
problem. The problem defines the needed solution, and understanding
the problem therefore reveals the required solution.
250 Object Thinking

Systems and Artifacts (continued)

Object thinking and structured analysis share a common belief that

the existing system must be understood before one proceeds with devel-
opment. They differ, however, in a very significant way. Structured thinkers,
like software engineers, tend to think of the system in terms of a bounded
artifact, a piece of software executing on hardware. Object thinkers view
the system as The System: the complex whole of the domain and all its
parts (objects), explicitly including all the human objects and all of the
patterned interactions of those objects. It is this System that we seek to
understand before attempting to intervene with any development project.
Understanding an entire System is seemingly a daunting task, one
reason why most software engineering texts suggest a much narrower
focus. Paradoxically, expanding our focus actually provides significant
benefits and has the effect of simplifying our work. The dictum “Every-
thing is an object” provides us with a single metaphor/model and a means
of partitioning even the most complex domain into a relatively small num-
ber of objects with easily understood behaviors.
This perspective also has the effect of redefining what is meant by
artifact and application as those terms are used in development projects.
Artifacts are objects or methods—nothing more. Applications are simply
scripted assemblies of objects, some objects being software simulations,
others being physical entities, and still others being human roles.
A side effect of this perspective is a kind of minimal-intervention
principle. Recognizing that introducing any change into a complex system
will have widespread, frequently unforeseen, and too often deleterious
consequences, we seek to minimize those effects by making the smallest
possible change. Examples of applying the minimal-intervention principle
include introducing a single new object, adding a new behavior to an
existing object, reassigning a behavior from one object to another, simu-
lating as faithfully as possible an existing object with software, and assem-
bling and scripting the least number of objects necessary to fulfill a single
story, as XP defines user stories.
 Chapter 9 All the World’s a Stage 251

We do need to understand more about the application—the context in

which our objects will perform and the precise tasks we expect our objects to
perform—before proceeding with development. Specifically, we need to deter-
mine the relationships, constraints, and interactions that will be required of our
objects in each specific circumstance. (Although we might discover new things
about our objects while engaged in our study of the application and might be
required to modify the intrinsic nature of our objects as a result, it’s a major mis-
take, from an object thinking perspective, to define objects in terms of an appli-
cation instead of a domain.)

Static Relationships
There are an open-ended number of possible relationships among objects.
Some are so common as to have special notation invented for them in modeling
tools (UML, for example): using, is-a-part-of, is-a-kind-of, and contains are
examples of common relationships.
From this large set of possible relationships, only two tell us anything
about the intrinsic nature of our objects. All of the others provide information
about the context—the application—in which our objects operate; they are sit-
uational relationships. The two relationships that describe intrinsic aspects of
objects are is-a-kind-of and collaborates-with.

Is-a-Kind-of Relationship
Classes are not essential to object thinking, merely a convenience. From a cog-
nitive standpoint, the use of classes allows us to think about an entire group
instead of all the members of that group—something we have been doing
throughout this book.
Combining the idea of classes with the is-a-kind-of relationship yields a
taxonomy, a particular kind of structural organization, that can then be used as
a kind of index to help in finding an object with a particular set of behaviors.
Figure 9-1 is a class hierarchy diagram—specifically, a taxonomy of the classes
in the mortgage trust example introduced in Chapter 7.
252 Object Thinking

Object

Form Valued
Collection

DataSource Investment
Interface

Collection Mortgage

LIMT ValueHolder

Mortgage
Application Payment Self EntryField
Evaluation

Expense Investment
Calculation
Collection Collection
Element

Employer Mortgage Payment

Collection Collection
Couple Person

Bank Home

Insurance Assessor
Agent

Figure 9-1 The is-a-kind-of relationships among the objects (via their
F09MQ01

classes) in the mortgage trust example.

Using a taxonomy requires only a general notion of the behavior or

behaviors you need in an object as a starting point. You can find an object with
that (or some of those) behaviors and then move up or down the taxonomy to
find an object that matches your needs in a more precise manner. If no existing
class of objects meets your needs, you can still use the taxonomy to simplify
your creative work by finding the object that best matches your needs and say-
ing that your new object is-a-kind-of the one you found. You then create a sub-
class and give it the precise behavior you want, and you use, via inheritance,
behaviors that already exist in the superclass.
From a cognitive point of view (we will discuss an alternative point of
view in just a minute), your taxonomy must follow two rules:
 Chapter 9 All the World’s a Stage 253

■ Exhibit a single line of descent (single inheritance) based on nothing

but the behaviors of the objects. Ideally, you should create your tax-
onomy (class hierarchy) as soon as you have completed side 1 of the
object cube. That way, you are not tempted to create a hierarchy
based on object internals, that is, methods or instance variables. Sin-
gle inheritance is important to avoid “abominations” that break the
taxonomy—the way the platypus breaks the Linnaean taxonomy of
fauna. (Linnaeus noted certain attributes of animal species such as
“warm blooded,” “lays eggs,” “live birth,” and “covered with hair”
and used them as the basis for classifying fauna. The platypus, a
furry mammal that lays eggs, breaks the classification rules and is
therefore an “abomination.”)
■ Subclass only by extension—by adding behaviors. A subclass should
be able to do everything its superclass (or superclasses) can do, plus
at least one more thing. This means that you can substitute an object
from a subclass in any situation calling for an object from its super-
class and still get the behavior you expect. If you define a subclass
by subtraction (taking away a behavior) or by redefinition (overrid-
ing a behavior), you create cognitive dissonance.1 Users of your tax-
onomy can never be entirely sure of the behavior they will get from
classes without looking inside those classes to see what they really
do. This is, of course, an egregious violation of encapsulation.

Class hierarchies can be viewed from an implementation perspective as

well as a cognitive perspective. Unfortunately, these two perspectives are not
always in concert. When they disagree, it is almost always because the hierar-
chy is perverted to conform to constraints imposed by a nonobject machine or
virtual machine (programming language).
In Chapter 5, “Vocabulary: Words to Think With,” it was noted that one
aspect of a class was to act as a repository for code and information shared by
all instances (objects) of that class. In terms of efficiency, this is a good idea
because storage space is minimized and changes can be made in a single place.
It becomes very tempting, however, to use this fact to justify creating your class
hierarchy based on shared code instead of shared behaviors. A similar error
arises from treating object variables (instance variables) as if they were data
attributes and then creating your hierarchy based on shared attributes. Always
create hierarchies based on shared behaviors, side 1 of the object cube, only.

1. Polymorphism—allowing different objects to respond differently to different messages—allows you to

create classes that override (replace) a superclass method with one of their own. There are reasons for
doing so, but they still create cognitive dissonance and make use of the objects containing overrides
more difficult. There are better ways to achieve the same goal—defining template or interface classes,
for example.
254 Object Thinking

Collaborates-with Relationship
Collaborations arise if, and only if, object A must use a service from object B
during the interval defined by object A receiving a message (request for service)
and object A returning the appropriate result object to the sender of the mes-
sage to object A. All collaborations are synchronous in the sense that object A
must invoke and wait for results from object B before it can complete whatever
work is required in response to the message it received.
Collaborations are almost always hard coded: they require manifest refer-
ences in the algorithm of the method where the collaboration occurs. Because
of the hard-coded nature of collaborations, we assert that they define part of the
intrinsic nature of an object. In every circumstance, in every context, the collab-
oration will occur if object A receives the indicated message.
Collaborating objects are very tightly coupled. For this reason, collabora-
tions should occur inside the encapsulation barrier, with objects occupying
instance variables, objects being received along with messages, and objects
occupying temporary variables. Collaborations with objects occupying global
variables are good or bad depending on the justification of need for a global
variable. Collaborations outside the encapsulation barrier are sometimes neces-
sary but should be avoided (by refactoring the distribution of responsibilities
among objects) whenever possible.
Wirfs-Brock introduced a dedicated model—a collaboration graph—that
was seldom adopted by actual developers. Figure 9-2 shows an example of
such a graph. The graph depended on the use of contracts (side 3 of the object
cube) to define potential connection points for potential collaborations. Deci-
phering the graph was difficult because of line tracing from client to server. Col-
laboration graphs (probably unintentionally) provide a visual metaphor
consistent with the object = software integrated circuit introduced by Brad Cox.
Interesting collaboration graphs look like circuit boards when viewed as a
gestalt.
 Chapter 9 All the World’s a Stage 255

31 32 19 5

View Outline
Element Element
18

34 35

31 32 33
19
5
Outline
Galley
View
View
Displayable
Printable View Object Document
Document Subsystem
Text
View
View

D
Text Graphic
Element 9

Document
Variable Element 10

3 Date Text 26

30 Time Paragraph 22 E

Page Graphic
20 12
Number Paragraph
1

Text F
27
Text Paragraph
G
Character
16 14
H
Heading
16 Link

Composable Object I

Document Display 4

6 7 8
Frame

A B

17 11 29 2

Marker Document C Text Composition

Manager Style Context
11 4

Figure 9-2 A visual model of collaborations, based on a graph in

F09MQ02

Designing Object Oriented Software, by Wirfs-Brock et al., 1991.

256 Object Thinking

Situational Relationship
Collaborations are not the only example of “uses” relationships among objects.
It can be said that object A uses object B if object B has asynchronously pro-
vided information or services to object A that object A finds useful for its own
work. For example, an airplane has a responsibility to identify itself, which
requires an id object, which leads to the assumption that at some point in the
past the airplane asked some national naming authority for an id object. That
request for an id object is an example of a uses relationship between the air-
plane and namingAuthority. (It is not a collaboration because the airplane does
not interact with namingAuthority each time it receives the id request.)
Except for collaborations, the fact that particular objects use the services of
others reflects the demands of the application (the situation) and not intrinsic
needs of the objects themselves. The same thing can be said of the particular
objects that are aggregated into components: the aggregation is a reflection of
the demands of a particular circumstance, nothing more. This is not to say that
it is unimportant to understand (and probably model) situational relationships,
but it is important to recognize, and keep in mind, that you are not modeling
the intrinsic nature of any objects.
What then do you need to understand about a situational relationship
application?

■ The objects that are participating in the work of the application

■ The objects that have a need to communicate with others and the
general nature of that communication
■ The “data”—really, the collective memory of objects involved in the
application; any relationships among data items
■ An overview, or gestalt, view of the application and its participants

The following sections describe models that might be helpful as a guide to

thinking about a situational relationship application.

Static Relationship Model

A static relationship model (the UML equivalent is a class relation diagram or an
object relation diagram) can provide all of this information. Static relationship
models (SRMs) are models of applications and not domains. They are called
static because the information portrayed in this type of model is supposedly
invariant. It must constantly be remembered that invariant applies to the time
and space boundary defined by each individual application; the relationships
might or might not exist in any other application. This is a major point of dif-
ference with data-driven approaches to modeling objects.
 Chapter 9 All the World’s a Stage 257

A data-driven approach assumes that objects are equivalent to data enti-

ties. Data entities are expected to be defined at the level of the domain because
it is assumed that they are the basis for creating a database that will be common
to all applications. This is one more example of perpetuating the traditional
view of software as data structures plus algorithms. A data-driven approach to
objects posits a singular database containing all of the data structures and
allows multiple applications—application is equal to algorithms only—to share
a common data structure.

Note Data modeling and object modeling are so dissimilar that a

named concept, the impedance mismatch problem, was coined to
reflect the difficulty of integrating object and database applications.
Some of the problems were simply technical and have been more or
less resolved. The inability of classical databases to generate
“changed events” or otherwise notify objects of a need to update
themselves when a cell in a database table has been altered is one
example. The basic difference, however, arises from the fact that
objects are domain-centric models, while schemas are machine/data-
base models. The effort required to “normalize” a conceptual data
model into a logical data model (one capable of implementation in a
database) essentially has to be reproduced in some degree each time
an object application wants to interact with a database—usually for
persistence. Immense amounts of effort have gone into the develop-
ment of middleware that allows objects and databases to interact with
each other.

An SRM, from an object thinking point of view, is a gestalt view of the

objects participating in an application (cast of characters) and relationships that
exist among those objects when in that application. Relationships most fre-
quently denote potential interactions among objects, just as a roadmap shows
potential travel2 among cities connected by those roads. Relationships might
also denote constraints on object behaviors or associations while the objects in
question are participating in a given application.
Most application SRM diagrams are little more than semantic nets (see
Chapter 6) with more precision, more detail, and an aura of specification

2. Roads connecting cities are obviously real. But the existence of a road does not mandate use by
vehicles; traffic remains a potential, not a certainty.
258 Object Thinking

instead of discovery. A semantic net also identifies the cast of characters—

objects—in a domain and reveals relationships among those objects. Semantic
nets were used as a heuristic starting point for object discovery. SRMs are
intended as a kind of specification: information to guide the developer in the
creation of the application.
Most of the information depicted in a semantic net will be used to define
and create objects. Some of that information will be used to define applications,
especially constraint- or rule-based information on the semantic net. Consider
the mortgage trust example discussed in previous chapters. Figure 9-3 shows a
fragment of the semantic net dealing with relationships between the price of a
home and the median price of homes in a neighborhood. The semantic net
fragment shows a specific business rule: the home price must be less than or
equal to the median price of homes in the neighborhood. This rule is not a
domain rule; it is an application rule. It should not become a basis for defining
the home, neighborhood, or price (actually money) objects because it does not
reveal the intrinsic nature of any object. Instead, it shows a relationship that
exists only in one or more applications (contexts), perhaps the “collect and val-
idate loan application data” application.

Figure 9-3 A semantic net fragment, showing a constraining relation-

F09MQ03

ship among objects in a domain.

Figure 9-4 shows an SRM for the mortgage trust example. Notice that it is
very similar to the semantic net: most of the same players are still present, and
most of the links among them are still evident as well. It differs from the seman-
tic net in the use of implementation names for many of the objects (classes), it
shows some of the interesting details of object construction (instance variables
and method names), and it uses terminology more suitable for implementation
than discovery. (Technical rather than domain jargon is allowed on an SRM but
not on a semantic net.)
 Chapter 9 All the World’s a Stage 259

accesses
Employer

accesses
Bank

completes accesses
Couple Assessor
uses DataSource
Interface
contains
Mortgage EntryField uses
Collection SelfEvaluating
accepts Rule

is-for-a
funds has Payment
LIMT
Mortgage Collection
has
contains
is-for-a
Investment
Collection uses
Payment Calculation
Home
Element

Collection Mortgage Form

Collection

Person Insurance Object

Agent

Figure 9-4 SRM depiction of the classes and relationship in the mort-
F09MQ04

gage trust example.

Collective Memory Map

One of the important responsibilities that an object might assume is to portray
and maintain some specific bit of information (data). Examples include num-
bers, integers, floating-point numbers, dates, times, characters, strings, and
money. None of these objects limit themselves strictly to the portrayal and
maintenance of information; they also have other interesting behaviors appro-
priate for the kind of information they maintain. Dates, for example, can per-
form calendar-related calculations and comparisons. Numbers can do
arithmetic. Strings have many behaviors similar to those of collections: iteration,
insertion, deletion, and arrangement, for example.
Given their specialized nature, data depiction objects frequently are found
occupying instance variables of other objects. An airplane object, for example,
has an instance variable named id, which is occupied by a string object.
Given the information-centric nature of the contemporary world, numer-
ous constraints and relationships exist between and among data depiction
260 Object Thinking

objects. Although it is possible to capture this type of relationship in a standard

SRM, doing so obscures, somewhat, the real nature of the relationship. Sup-
pose, for example:

■ An airplane object has instance variables for id, containing a serial

number such as N543UE; transponderCode, a four-digit integer; and
status, containing a character whose value indicates that the airplane
is in normal operation, hijacked, in a declared emergency, or in con-
trolled airspace.
■ An airspace object has an instance variable, assignedAircraft, that
contains a collection of blips.
■ A blip object has instance variables for id, transponderCode, altitude,
latitude, longitude, vector, priority, and status.

Figure 9-5 shows a fragment of an SRM that relates the three objects under
discussion. This fragment obscures a great deal of detail about the actual rela-
tionships involved—for example, the following rules:

■ Transponder code value must equal 1200 for VFR flight.

■ Transponder code value must equal 7500 if plane has been hijacked.
■ Transponder code value must equal 7700 if plane is in a state of
emergency.
■ Transponder code value must equal 7600 if plane has lost communi-
cations.
■ Transponder code must equal value specified by ATC for its blip rep-
resentation.
■ Transponder code value must be consistent with airplane’s status
value.
■ Blip ID and airplane ID must be consistent.
■ Blip priority must be 1A if transponder code is 7500, 7600, or 7700.
 Chapter 9 All the World’s a Stage 261

Figure 9-5 Relationships among airplane, blip, and airspace.

F09MQ05

It would be possible to model all of these constraints simply by adding

relationship symbols between blip and airplane on the SRM—if you are willing
to sacrifice a great deal of readability. Alternatively, since all of these relation-
ships are among objects occupying instance variables of other objects, there is
value in modeling them directly in a separate model, a collective memory map
like that depicted in Figure 9-6.

Figure 9-6 Relationships among objects occupying instance variables

F09MQ06

of other objects in the airplane, blip, and airspace example.

262 Object Thinking

A collective memory map (CMM) is another type of static diagram con-

structed according to the following rules:

■ Objects appearing on a CMM are occupants of instance variables in

other objects.
■ Objects appearing on a CMM must have a primary role of data depic-
tion and maintenance (maintenance implies self-validation, self-per-
sistence, and self-based access control); they can be considered
primitives in the same sense that a database defines certain objects as
primitives. Most often these include numbers, characters, strings,
dates, money, time, and similar objects.
■ A kind of collection that is restricted to containing “primitive”
objects, called a ValueHolder, can also be depicted in a CMM.
■ Each object depicted is named using dot notation that identifies the
container object name and instance variable name—for example,
customer.lastName.
■ In those cases in which the instance variable contains a ValueHolder
instead of a primitive, the dot notation is extended to include the
label (key) associated with the actual primitive—for example, cus-
tomer.description.gender.
■ Relationships among objects on the CMM are drawn as connecting
lines with appropriate labels.

Figure 9-7 depicts a collective memory map for the mortgage trust example.
Collective memory maps are situational: they reflect constraints that exist
among “data” objects in the context of a particular application. In some cases,
it might be useful to create a domain-level version of a collective memory map.
A domain-level collective memory map would have some things in common
with an enterprise memory model, namely a global depiction of all the “data” in
the domain and the objects in which it might be found and any relationships
among those objects that were truly invariant across the domain. If your enter-
prise is interested in constructing such a global model, you need to remember
some important caveats:
■ Object thinking presumes that “data” objects represent information
or knowledge that objects need to perform their assigned tasks,
while enterprise data models presume to depict all the data that the
“system” must remember about objects. This is a critical distinction.
 Chapter 9 All the World’s a Stage 263

■ The global model should contain only relationships and constraints

that are invariant across the domain.
■ A CMM for an application can add, delete, and modify relationships
to reflect the situational context. If there is a conflict, the global map
must be modified.

Employer.weeksWorked MortgageApplication.status
Mortgage.status

Employer.wages
Investment.current.Value
LIMT.amountOnHand
Investment.id Neighborhood.medianPrice

Person.employer EntryField.heldClass must be<=

Person.bank Home.salePrice
Bank.customerRating

Couple.marriageDate Application.Contents
Home.location

contains contains

Application.contents.home

matches
Application.contents.marriageDate Home.legalId

Figure 9-7 A collective memory map for the mortgage trust example.
F09MQ07

Architecture
Almost all applications have an overarching structure or means of organization,
which might be called architecture. Architecture affects the way you think
about implementation, especially when the architectural pattern is so common
as to be assumed. For example, traditional structured development presumed a
hierarchical command-and-control architecture as reflected in a program struc-
ture chart (Figure 9-8). This visual pattern is seen in the mainline-plus-subrou-
tines source code organization scheme (Figure 9-9). The hierarchical control
architecture is a codification and expression of a lot of ideas about “good” pro-
gram design, ideas that support most of the rest of structured development
approaches.
264 Object Thinking

Master Control
Module

Afferent Modules Transform Modules Efferent Modules

Figure 9-8 The proper organization (architecture) of a program consis-

F09MQ08

tent with structured development ideas.

Initialize procedure
Main routine
Input data routines
conditional flow logic
end input routines
Process data routines
case statement depending on data types
end process routines
Output routines
conditional flow logic
end output routines
end main routine
Termination routine

subroutine one
transform code
end subroutine one

subroutine two
transform code
end subroutine two

Figure 9-9 A structure depicting how actual source code might reflect
F09MQ09

the hierarchical control architecture.

A prototypical architecture that reflects object thinking ideals and princi-

ples also exists: model-view-controller (MVC). MVC is not a mandated architec-
ture. You can use almost any of the commonly encountered architectural
patterns (see the sidebar “Architectural Patterns and Objects”), but an investiga-
tion of MVC is instructive and important for the way it illustrates and rein-
forces the other aspects of object thinking that have been discussed in
previous chapters.
 Chapter 9 All the World’s a Stage 265

Architectural Patterns and Objects

MVC is but one of a number of architectural patterns or prototypical ways
of organizing a group of objects engaged in a collective task. Commonly
encountered alternative architectural patterns include hierarchical control,
pipes and filters, client/server, and several variations of the blackboard
architecture. With the exception of hierarchical control (the antithesis of
object ideals), all of these alternative architectures can be consistent with
the principles of object thinking.
Regarding pipes and filters, readers familiar with UNIX/Linux will
immediately recognize this pattern. Two kinds of objects are involved:
pipes, which are connectors and possibly temporary storage locations;
and filters, which are places where specialized services (transformations)
are performed. Figure 9-10 illustrates a small pipes-and-filters architecture.

Figure 9-10 Objects move along the route provided by the pipes and
F09MQ10

obtain services from the objects embodying the filters.

Traditional thought about pipes-and-filters architecture reflects tradi-

tional thinking about the separation of data and process. From that non-
object point of view, data passes along the pipes (pushed or pulled by the
connected filters) and is manipulated by processes at each filter station.
Simply reversing the idea about who is in charge and assuming that
everything is an object makes this a perfectly acceptable architecture for
use by object thinkers. The architecture merely provides transport objects

(continued)
266 Object Thinking

Architectural Patterns and Objects (continued)

(pipes) and service providing objects (filters). A large pool of objects

requiring a similar set of services can ask the pipes for transport to the filters
whose services they need, take advantage of both, and effect self-processing
in a very efficient manner.
Client/server architectures are the mainstay of the networked business
world. Two-tier client/server systems are almost identical with MVC, with
the model (usually a database) physically located on a server and most of
the view and control aspects on clients. Conversely, MVC can be seen as
simply an n-tier (or peer-to-peer) client/server architecture. Object thinking
will frequently, and naturally, lead to the construction of a client/server
architecture, the only difference being the overall distribution of responsi-
bilities and the absence of overt controller, manager, and scheduler objects.
Blackboard architectures actually come in at least three variations,
which I’ll label bulletin boards, blackboards, and whiteboards. All variants
have common elements as illustrated in Figure 9-11: service providers,
requesters, a common communication space, requests for service, and
results. Each of these elements is, of course, an object.

Figure 9-11 Elements common to all blackboard systems.

F09MQ11
 Chapter 9 All the World’s a Stage 267

Architectural Patterns and Objects (continued)

In the simplest form of blackboard architecture, the bulletin board,

requesters post requests in the common space, and service providers con-
stantly poll that space looking for requests they might service. Upon find-
ing an appropriate request, the service provider retrieves it and provides
appropriate satisfaction, a result, which is then returned to the common
space, where the original requester finds it (constantly polling the space—
“Are you there yet?”) and retrieves it.
Request and result objects are relatively straightforward containers—
of the information necessary to specify a result and the result that can be
queried by the requester upon retrieval of the information. A ride board,
such as that found in most university student unions, is a prototypical
example of a bulletin board system.
Give the common space some interesting responsibilities, and you
have a blackboard variant. The common space maintains a list of service
providers to which it can match a posted request. When a match is found,
the request is forwarded appropriately. The common space also retains a
list of requesters and their requests so that when a result is posted it can
forward the result to the appropriate requester. Care must be taken not to
turn these responsibilities into a command-and-control structure.
Whiteboards are similar to the tuple spaces proposed by David
Gelernter3 and implemented in his Linda programming language. In this
variant, the shared location—the whiteboard—is called tuple space and
the request and result objects are called tuples. Tuples can actually per-
form work, transform themselves by combining with and interacting with
other tuples, in tuple space. The tuple space itself is also given enhanced
responsibilities for filtering and feeding forward (to the next hierarchical
level of tuple spaces) its contents (tuples).
All of these architectures are perfectly consistent with the ideals of
object thinking as long as the assumptions about control, distribution of
responsibility (factoring), and object autonomy illustrated in MVC are
respected.

3. Gelernter, David. Mirror Worlds. Oxford University Press, Reprint edition (January 1993).
268 Object Thinking

Figure 9-12 depicts the model-view-controller architecture. The figure is

actually a slight variant of the architecture proposed as the Smalltalk language
and development environment was being created at Xerox PARC. MVC is almost
always discussed in the context of a graphical user interface environment—
invented simultaneously with Smalltalk—but the important principles behind
the architecture are not limited to such an environment.

Application
Application View
Dispatcher
Application Object
composite w/
w/ application
application’s
dispatcher
object-object
dispatcher

composed of
object views Object-View
Dispatcher

Composite
object view

Object-Object
Dispatcher
Object-View
Dispatcher

Object View

Object-View
Dispatcher

Figure 9-12 Generalized model of the MVC concepts.

F09MQ12

MVC is grounded in the idea of specialization and distribution of

responsibilities—specifically, the identification of three main categories of
responsibility (presentation or visualization, computational work, and coordi-
nating communications) and assigning those to different sets of objects.
The model comprises those objects actually engaged in accomplishing the
work objectives of the application. Model objects have a need to communicate
with one another by sending messages and by notifying other objects of inter-
nal state changes. In Figure 9-12, model objects are depicted as round circles.
 Chapter 9 All the World’s a Stage 269

One of the model objects assumes the role of the application object. The
application object assumes responsibility for startup and shutdown activities in
collaboration with the other model objects and acts as a kind of global (within
the application) repository of objects and information that need to be visible to
other model objects while the application is running.
The view consists of a hierarchically organized collection of objects—for
example, widgets, windows, and icons—whose primary task is creating visual-
izations of objects so as to be comprehensible to human beings. This definition
presumes a GUI environment and reflects the history that shaped the original
MVC architecture.
A view object does not have to be a GUI element. If an object creates a rep-
resentation of itself other than its native implementation, that too would be a
view object. Examples of nonsensory views would include bit stream serializa-
tions that allow objects to exist in a relational database or an XML statement that
allows an object to be shared across applications and implementation languages.
In Figure 9-12, view objects are depicted as rounded rectangles. The hierar-
chical organization of views reflects the fact that simple objects have simple views
and more complicated objects have views that are collections of the views of the
simpler objects that they comprise. It’s also possible for any object to possess
multiple views of itself. An integer, for example, might have a simple bitmap
view that depicts the value of the integer, and it might have an update view
consisting of a widget depiction (rectangular entry field area) plus the bitmap of
its value. Figure 9-13a illustrates the hierarchical composition of a complex
view, and Figure 9-13b provides an example using the mortgage application.

Object1 View
Application Object

Object3 Object1
view
Object2
composite view
Object4 Object3
view

Object4

Object2
Application View

Figure 9-13a Hierarchical composition of a complex view.

F09MQ13
270 Object Thinking

Figure 9-13b The application completion screen in the mortgage trust

F09MQ13

example is a hierarchically organized collection of object views.

The controller is badly misnamed. Control in the classical programming

sense is anathema to object thinkers and right-thinking autonomous objects
everywhere. I’ll substitute the word coordinator for the rest of this discussion.
Coordination refers to the tasks involved that allow one object to send a mes-
sage to another and for an object to notify any interested object of a state
change in itself.
Ensuring that it’s possible for two objects to exchange messages is a matter
of visibility: the objects have to be able to see one another. An example of coor-
dinating visibility is the application object having a variable that contains a col-
lection of all the objects participating in the application. Should one object need
to talk to another, it can ask the collection for an object by name or by criterion.
Similarly, the application object might have a variable that will contain the cur-
rently active or currently selected object, to which other objects might need to
send messages. Because the application object is global within the application,
all other objects can send messages either to it or to the objects occupying its
instance variables.
 Chapter 9 All the World’s a Stage 271

Event (state change) notification was the main responsibility of historic

controller objects, another manifestation of the GUI-centric history of MVC. We
will use a special-purpose object, an eventDispatcher, to provide all the neces-
sary coordination associated with event notification. Dispatchers were intro-
duced earlier; the following paragraphs will serve as a reminder.
An eventDispatcher is a simple two-dimensional table, the first column of
which contains a list of events that can be dispatched and the second column
of which contains a collection (possibly an ordered collection) of notification-
Request objects.
A notification request is a dyad consisting of the name/ID of the object
requesting notification of an event’s occurrence and a message to be sent to
that object as means of notification. If the notification collection is ordered (for
priority, for instance), the notification request can be a triad of name, message,
and priority number.
An eventDispatcher can add or delete events, must accept and delete reg-
istrations on request, and, when an event is detected, must tell each registration
in the associated queue to execute itself (send the contained message to the
contained object reference).
If the registrationQueue is an ordered collection, it must reorder itself
each time a registration is added or deleted.
The eventDispatcher objects in Figure 9-12 are depicted as rectangles
(long axis is vertical). Each object has a small solid triangle inside itself, indicat-
ing that every object has the capability of accepting registrations for changes
within itself (as indicated on side 6 of the object cube).
Rectangles appearing between an object (circle) and a view (rounded
rectangle) coordinate and ensure consistency between the object and its view.
Just as views are composites, so too are these event dispatchers. A composite
object (Customer, for example) might have a composite view consisting of the
views of each of the objects occupying an instance variable of Customer (the
strings in fName, lName, and MI, for example). Although it’s possible for each
object participating in this composition to use its own individual dispatcher to
coordinate changes, it’s also possible to create a single dispatcher and have
each object add an event line to that dispatcher as it adds its view to the com-
posite. Figure 9-14 illustrates this composition.
272 Object Thinking

Customer
Object
widget3 registration

Customer View Event Registration Queue

Obj1 View
Cust-E1
Cust-E2
Obj1-E1
Obj2-E1
Obj2 view Obj3-changed [Widget3 update]
Widget3-changed [Obj3 update]
Obj4-changed
Widget4-changed

Obj3 registration Obj1

Obj3 View Obj4 View Obj5 View Obj6 View Obj3

(as widget) (as widget) (as widget) (as widget)
Obj4

Obj2
Obj5

Obj6

Figure 9-14 The composite object, Customer, might use a composite

F09MQ14

dispatcher instead of each object using its own dispatcher.

In addition to the explicit elements of MVC, we also assume an outside

world consisting of all the objects outside the application object. This is the
realm of human being objects, legacy system objects, and operational platform
objects such as hardware, operating systems, mouse managers, and keyboard
managers. (Of course, any object whose name includes the term manager is
immediately suspect and should be examined carefully for violations of object
thinking principles.) Although it’s true that everything is an object, it’s equally
true that many of the things in this realm operate in a very nonobject fashion.
From the perspective of the application object and the objects making up its
cast of characters, the outside world is a source of information and events. It’s the
presumed destination of results, but whether someone actually uses the results an
object was asked to provide is not of much concern. Objects live to serve. Some-
one on the outside asked for a service, and it was provided. The object assumes
 Chapter 9 All the World’s a Stage 273

that the client knew what it was asking for and had a use for the result. This is an
anthropomorphic way of saying that objects are not and should not be aware of
their clients, even when those clients are not other software objects.
The application object, or one of its cast members, might require services
from objects in the outside world. In that case, it might need to display itself
using an update view, an implicit request to an outside world object to provide
information by altering the update view. In other cases, it might need to find
and send a message to an object in the outside world, usually by sending a
message to the operating system object via the application object.
A composite eventDispatcher associated with the application object also
handles events generated in the outside world of which objects inside the appli-
cation (or the application object itself) might need notification. In Figure 9-12,
the application dispatcher has three parts (view-object coordination events, out-
side world events, and global notification events) illustrating its composite
nature.
The MVC architecture has been criticized (in some cases by those who
created it), but it is an excellent example of how to organize and coordinate a
group of objects without imposing any kind of centralized control. The distri-
bution of responsibilities and the creation of special-purpose objects such as
the dispatchers allow complete flexibility, even at run time. Dispatcher events,
for example, can be added and deleted by sending a run-time message, as can
registrations for those events. It’s even possible to dynamically configure the
event registration dyad at run time. MVC provides an existence proof of how
the ideals of object thinking can be realized.

Dynamic Relationships
All of the static relationships (except class hierarchy and collaboration) can be
seen as “setting the stage” information. They determine the scenery and props,
the marks that actors use to navigate the stage, the cast of characters, and an
outline of the plot. A script actually determines what happens on the stage dur-
ing the performance of the play. A script is simply a collection of messages sent
from actor to actor, interpreted by receiving actors, and resulting in subsequent
messaging. On occasion, actors respond to cues other than messages—that is,
signals, such as the ringing of a phone, which is the phone’s way of notifying
other objects of a change in its state.
Capturing and modeling the dynamics of an application (following the
stage metaphor) consists of detailed modeling of scripts and event notification:
the two sources of “cues” for our participating objects.
274 Object Thinking

Scripts
During discovery, we told stories about object interactions. Some of those sto-
ries we wrote down on cards; others we modeled with interaction diagrams.
Most of those stories exhibited a nested structure, analogous to the way a play
is broken up into acts and acts into scenes. Design and implementation activities
require that we turn those stories into specifications, program code, or both.
The same tool, an interaction diagram, is used to discover conversational
requirements and specify a script. The only difference will be in the labels
attached to objects, messages, events, and returned objects. During discovery
we want to use, exclusively, domain language in describing our stories. For
design purposes, our diagrams should have labels that correspond to actual
classes and actual messages as they appear in message protocols recorded on
the object cubes. (In the Behavior! tool mentioned earlier, interaction diagrams
have color-coded labels: red meaning that the label has not been mapped to an
implementation term and reflects user vocabulary, black for user vocabulary
that has been mapped to an implementation term, and blue for the implemen-
tation term. The diagram itself does not change, only the labels, assuring con-
sistency between our domain and the simulation of that domain we are creating
in software.) Figure 9-15 depicts the discovery and script versions of a conver-
sation from the mortgage trust example.

Figure 9-15 The diagram on the left has labels reflecting domain vocab-
F09MQ15

ulary. The one on the right is the identical conversation with labels reflect-
ing implementation vocabulary.

Customers, working with developers, spend time factoring their stories

into discrete cards with the goal of creating cards that can be estimated (effort
required to realize the story in code) and that have a discrete focus. This may
 Chapter 9 All the World’s a Stage 275

require refactoring an ambiguous or overly ambitious story. No intermediary

design is required. An XP user story maps to a discrete bit of functionality of a
larger program or system and should probably involve fewer than six objects
exchanging ten or so messages. Purely as a thought-organizing aid, even XP
developers often take the time to sketch a rough interaction diagram as a means
of factoring stories and as a prelude to coding. Although not one of the official
twelve practices of XP, a quick design session using a whiteboard clearly has a
role in clarifying the intent of the story and outlining the participating objects
and their interactions.
For example, consider the story card “Application form is completed and
validated,” a story that almost certainly needs to be factored. Figure 9-16 (a–c)
shows rough sketches of interaction diagrams that might have been sketched
during a conversation4 among developers and the on-site customer. Drawing
the diagrams reveals ways to factor the original story. (The story might actually
stay intact, and only the tests and tasks completed by the programmers are fac-
tored to reflect the design implications of the interaction diagram sketches.)

Figure 9-16a The original story that, whenever it looks as though a lot is
F09MQ16a

going to happen, is factored with a placeholder for a substory inserted.

4. An XP story has three parts, card, conversation, and confirmation. The card is the tangible evidence
of the story and a place to record information about the story (estimate and priority, for example). The
conversation may be verbal or may involve the use of diagrams such as those depicted here. The con-
versation is intended to expand understanding and to reveal the design behind the story. Confirmation
is the acceptance test, written by the user and implemented by the developers, that proves the code
accomplishes the desired task.
276 Object Thinking

Figure 9-16b Substory: “Entry field obtains and validates contents.”

F09MQ16b

Figure 9-16c Substory: “Application form validates self.”

F09MQ16c
 Chapter 9 All the World’s a Stage 277

Note that the interaction diagrams created as part of the conversation

about the story expose a need for object interactions that detect and resolve
problems—error handling.
A final note on scripts: It’s quite possible, and sometimes desirable, to
implement scripts as first-class objects, as an ordered collection of messages.
When told to execute, the script sends the messages it contains, in order; accu-
mulates the results; and keeps track of its own status in terms of its successful
execution. This approach adds discipline to a conversation, ensuring that it
occurs as it was supposed to, without adding control. The script object encap-
sulates a conversational fragment, and it’s the script object itself that succeeds
or fails in its assigned task. As a first-class object, a script can be modified by
adding and deleting message objects without the need to recompile the appli-
cation containing those script objects.

Event Dispatching
Events are cues that can be used by the developer to script object interactions.
Side 6 of the object cube lists the events (cues) available for use. Our stories
and the interaction diagrams that might have been created as part of the con-
versation about those stories provide us the information needed to link an
event (cue) with resulting messages sent to the objects being cued. The object
cube tells us of the state changes an object deems of potential interest to others,
and the interaction diagrams note occasions when state changes affect the flow
of a conversation among objects.
The interaction diagrams also confirm an important aspect of object think-
ing: the only object directly aware of a state change is the object in which the
change occurs. The double-headed arrow used in an interaction diagram to
denote a state change signal must always be directed to the object in which that
state change occurs, never to other objects in the diagram. Figure 9-17 illus-
trates this important caveat.
278 Object Thinking

Figure 9-17 State change notification is always a “message” to oneself, never to

F09MQ17

other objects. A state change can trigger multiple messages to other objects,
reflecting the contents of the event dispatcher, as shown in the diagram on the left.

All event notification is performed via an event dispatcher located in each

object. Event dispatchers must be populated for an application to execute—that
is, each must contain a list of events and must have registrations5 for those
events consistent with the needs of the application as expressed stories (inter-
action diagrams). Because most objects will have little if any interesting state
behavior, populating an event dispatcher will be relatively straightforward.
However simple an object’s state behavior may be, it would be nice to
have some corroboration of the information on side 6 of the object cube and in
our stories. Taking the time to sketch a state model can provide this verification.
State diagrams are also useful for modeling (revealing) constraints on object
behavior (see the following major section, “Constraints”) and for modeling
event-driven (stimulus-response) scripts. A state chart, as devised by David
Harel, has become the modeling tool of choice in the object community.

State Modeling
Figure 9-18 shows a state diagram for a mouse and part of a diagram for a
mouseManager object. (mouseManager is a name reflective of a popular oper-
ating system environment—not of object thinking.) The dotted line in the

5. Some of these registrations will be made as part of the application initialization process—getting the
cast (objects) ready to perform—and others will come and go during the running of the application.
 Chapter 9 All the World’s a Stage 279

diagram reveals that a mouse really consists of two concurrently operating

objects, a movement detector and a button. (Although most mice have more
than one button, it isn’t necessary for the state chart to model them separately.
Scroll wheels and other mouse embellishments are not modeled here but
would be additional concurrent state machines within the mouse.) Both objects
have very simple state, but the diagram confirms the events that need to be
listed on the object cube for each object—click and out of order for the button;
newXYZ and out of order for the movement detector.

Figure 9-18 This Harel state chart depicts a mouse and its composite
F09MQ18

parts plus a fragment of a mouse manager.

Inclusion of the mouseManager diagram fragment illustrates how object

thinking affects distribution of work. Instead of making a complicated mouse
capable of generating events for all possible state change combinations, we
keep the mouse simple and have other objects assume responsibility for trans-
forming simple events into composite events—in this case, for generating a
double-click event. This example use of a state chart also reflects the XP com-
mitment to simplicity. By separating the generation and interpretation of mouse
events, we create objects that are simpler and more composable.
The partial state chart shown in Figure 9-19 shows partial state charts for
a mortgage application and an entryField. Assuming that the entry field in question
is one of those contained on the form, the diagram represents event registration
280 Object Thinking

information that needs to be accounted for and a constraint on the behavior of

the application itself:

■ Any change in the contents of an entryField object forces the con-

taining application into the nonvalidated state. This means that the
entryField object cube must contain a changed event and that the
application must register for notification of that event with all of its
contained entryField objects.
■ The application cannot be funded if it’s in the nonvalidated state,
reflected in the fact that the funded trigger operates only while the
application is in the validated state.

Figure 9-19 Partial charts for application and entry field, from the mort-
F09MQ19

gage trust example.

Unless the developer and customer are discussing an attempt to model a

complex event-driven composite object—a graphical user interface perhaps—
most state modeling is used only for the purpose of helping us think about pos-
sible constraints on an object’s behavior, to confirm that we have listed all nec-
essary events on each object’s cube, or to confirm/reveal the need for one
object to register for events (state changes) with other objects.
 Chapter 9 All the World’s a Stage 281

Constraints
We have a stage, actors, and a script (application context, objects, and a script
object plus populated event dispatchers); and the play (application) is ready to
open. Except … sometimes our actors have to obey rules or are otherwise con-
strained in their actions. We need to accommodate this need as well.
Objects, like people, can have behaviors that should not be exercised in
certain circumstances. All constraints are local and arbitrary (for example, the
rules of culture that allow one behavior in one place but not another), and
because of this, the modeling and the implementation of constraints are not
intrinsic to the definition of objects, only to the context (application) in which
those objects perform at any given time.
Classical thinking about object development made this difficult, primarily
because rules and constraints were primarily seen, and modeled, in terms of
static relationships between or among objects. Because these relationships
were supposedly static, they tended to be hard coded in object designs and
implementations.
With a few exceptions, object thinking suggests that constraints and rules
should be thought of as objects in and of themselves. We have talked in this
vein throughout the earlier chapters and will make the idea explicit in the next
section, “Self-Evaluating Rules.”
Constraints on objects can be direct, inhibiting a behavior (actually allow-
ing or prohibiting the execution of a method) or setting limits on the values of
objects contained in instance variables (the knowledge that objects work with
when doing their jobs). Constraints can be indirect, reflecting relationships
among objects instead of object internals.
Examples of direct constraints include

■ Inability to respond to a message because the object is already

engaged in performing work in response to a previous message.
■ Limiting the set of objects to which an object will respond when sent
a particular message. For example, the airplane would allow no one
except a manufacturer object to send the message that sets its ID
(serial number).
■ Restricting the values that can be assumed by an object—telling an
integer that it can assume only values between 10 and 100 would be
an example.

An indirect constraint is illustrated with the following example: a marriage

object must associate two, and only two, person objects. (Of course, this is a
local or cultural rule, not universal, and so does not reveal the intrinsic nature
of a marriage object.)
282 Object Thinking

Constraints are revealed at different times in the development process.

Stories, told to identify objects, are the most common source. Static models (dis-
cussed previously) reveal constraints that might have been assumed in the sto-
ries without explicit mention. Thinking about implementation reveals many
others, especially constraints on values that objects may assume and state-based
constraints.

Note The primary sources of constraints are the definition of stories

by the customer, the discussions of stories by the customer and the
developers, and the development of tests to demonstrate that the sto-
ries are correct. The artifacts discussed in this section (interaction dia-
grams and state charts) can be used during those three activities.

Implementing constraints must be done with program code. That code is

generally located in one of three different places: the affected method, a man-
ager/controller object, or a rule object (by encapsulation). The preferred option
is using a rule object. Hard coding makes objects brittle and mandates undue
maintenance if changes in constraints occur over time. Manager/controller
objects violate object thinking principles (and, from a purely pragmatic point of
view, result in unnecessary complexity). Creating a class of objects capable of
realizing any kind of rule—production rule, business rule, formula, and so on—
is the object thinking solution to the constraint problem.

Self-Evaluating Rules
A dependent is eligible for family coverage if she or he is under 18 years of
age, between the ages of 18 and 24 but a full-time student and receiving a
majority of her or his support from the insured, or if she or he is dependent
due to disability or illness.
Your interest rate is the average prime rate (as published in the Wall Street
Journal) for the 30 days preceding the issue date of your statement plus 4.5
percent.
An Employee may have 0 or 1 Spouse.
Py = (1/360 * FA) + (1/12 * (.075 * CB))
 Chapter 9 All the World’s a Stage 283

The preceding are examples of business rules. Businesses are rife with
these kinds of statements, formulas, conditionals, and mandated relationships.
Despite the ubiquity and importance of business rules, they are seldom
accorded the status of first-class objects. Instead, rules are documented and pre-
sented as static relationships among classes. (Traditional development methods
deal with rules in the same way—for example, using entity relationships in a
data model.) These static relationships are then supposed to be established and
enforced by appropriate code in object methods or by establishing special-pur-
pose objects, such as a database management object, that specialize in the
maintenance of relationships.
Object thinking mandates that “everything is an object.” A rule6 should
therefore be an object and, like any other object, have the ability to perform a
specific set of services: modify itself and evaluate itself. By modify, we mean
the rule must be flexible: adding, deleting, or extending clauses. By evaluation,
we mean applying computational rules to itself so as to return an object encap-
sulating a meaningful result—in the simplest case, returning a Boolean object
that encapsulates whether the rule was satisfied.
Applying object thinking to the problem of creating a rule object leads to
observations about structure and responsibilities.

Structural Abstraction of a Self-Evaluating Rule

A useful object thinking heuristic for decomposition is to take an occurrence of
such an object as it appears in the real world and separate its physical parts. We
can do this with a simple example of a rule expressed as a formula:
X = 4q + (p*r)
Analysis of this equation (rule) yields five distinct types of physical
components:

■ X, q, p, and r represent things that are currently unknown but know-

able. Following convention, we will call these items variables.
■ 4 represents a constant value.

6. Many years ago, I formulated the basic ideas of self-evaluating rules and lectured my classes on the
concept. I didn’t actually have a rule object implemented (leaving that as an exercise for the student).
Without implementation, self-evaluating rules were a nice idea, but … Kevin Johnson, one of my grad-
uate students, who later became a colleague and great friend, would occasionally complain that rules
could not be implemented as I was describing them (as described in this chapter). So one day we sat
down and implemented them in Smalltalk, with him doing almost all of the programming. The final
concept of self-evaluating rules is the result of Kevin and me discussing them over many months and
finally implementing them in at least one object language. He and I coauthored an unpublished paper
that is the foundation for most of what is said in this section. I want to thank Kevin and acknowledge
his invaluable assistance in finalizing the ideas in this section.
284 Object Thinking

■ + and * are behavioral operators.

■ ( and ) in combination represent an aggregation and precedence
operator.
■ = is an operator of symmetry, indicating that the thing on one side of
the equation is in some sense symmetrical with, equal to, or to be
assigned to the thing on the other side.

The order in which these components appear is important, as is the asso-

ciation of elements with one another (for example, determination of the
receiver/operator/argument relationship). Our initial, structural, abstraction of a
rule then follows:
A rule is an ordered collection of variables, constants, and operators.
Each of these components can be analyzed to discover its behavioral char-
acteristics. Constants are perhaps the simplest, most often being instances of
known classes such as integer, float, real, character, string, and so forth,
although nothing prevents them from being any known object with the ability
to provide a fixed or constant value.
Variables are more interesting. At first glance, we seem to have two differ-
ent types: those, like X, that equate (are assigned) to the resolution of the entire
rule; and those, like p, q, and r, that represent a discrete value that can be used
to resolve the rule.
If we think of variables as a place where a value (an object) could be but
currently is not, we can posit a structural abstraction for a variable. In the case
of variables like X, that abstraction would consist of a targetObject and a setter-
Message. For the q, p, and r variables, the abstraction is a sourceObject and a
getterMessage. Behaviorally, a variable is responsible for obtaining its value, for
instantiating itself. It does so by sending the getterMessage to the indicated
sourceObject.
In the realm of objects, we are accustomed to treating operators as mes-
sages. In our simple example, we see three nuances of operator: first, those rep-
resenting standard messages sent to objects to invoke behavior (+ and * being
examples); second, those that instruct the compiler/interpreter/virtual machine
to establish precedence; and third, assignment operators. The hardest part of
implementing self-evaluating rules was dealing with the relationship between
operators and the objects related to each. The solution turned out to be creating
a term class that had receiver, operator, and (optionally) arguments as instance
variables. This turned our rule into an ordered collection of terms, but the con-
ceptual essence of self-evaluating rules was preserved.
Our attention can now turn to behavior and the dynamics of the rule and
term structures we have defined.
 Chapter 9 All the World’s a Stage 285

Behavioral Abstraction
Side 1 (responsibility list) of an object cube for a self-evaluating rule would
have the following entries: modify self and evaluate self. The second responsi-
bility requires collaboration with the ruleElement objects, namely the constants,
variables, and operators.
Side 2 (description and stereotype) might show a stereotype of ordered
collection, reflecting the modify self responsibility coupled with the definition
of a rule: an ordered collection of variables, constants, and operators.
Side 4 (knowledge required) would indicate a need to know about the
ruleElement objects that need to be added, deleted, and so forth but, surpris-
ingly, little else.
Side 5 (message protocol) would define messages for the following: add
an element at a location, delete a designated element, evaluate, instantiate, and
resolve.
Side 3 (contracts) would show that evaluate, add, and delete messages
would be public, while instantiate and resolve would be private.
All basic behavior is now defined. A rule is asked to evaluate itself. It, in
turn, sends itself the instantiate message, causing itself to iterate across its ele-
ments and ask each variable object to instantiate itself (turn itself into an actual
value by sending its message to its receiver). Once all variables are instantiated,
the rule then sends itself the message to resolve—apply the operators appropri-
ately and return the result (with assignment taking place as needed).
Rules are recursive. Any element of a rule can be replaced with a rule.
That’s more than enough flexibility and power to get yourself into trouble
unless you use other object thinking principles, such as simplicity and local
action, to eliminate the need for truly complex rules.
Every object can be given the responsibility to validate itself simply by giv-
ing it a collection of rules to use as collaborators in that process. Those rules
can be modified at run time, just like any other object, merely by sending
appropriate modification messages. Variables can be context sensitive in their
instantiation.
This type of self-evaluating rule was presumed in examples in previous
chapters. The form example assumed that every entry field would have a col-
lection of self-evaluating rules that it would use to validate its contents.
Figure 9-20 provides three examples, the first two showing simple entryField
validation rules and the third showing a more complex rule for form validation.
286 Object Thinking

RETURN "True" IF (objectEntered WhatIsYourClass) EQUALS "integer"

operator op op
constant variable const

RETURN "True" IF (objectEntered yourValue) < "100"

op op op
const var const

RETURN "True" IF "ZipCodeDirectory" CONTAINS

op op op
const const

((widgetNamedZipCode yourValue) variable

CONCATENATEDWITH op

(widgetNamedCity yourValue)) variable

Figure 9-20 Sample rules with identification of element types.
F09MQ20

Creating rules as first-class objects is essential to eliminating the kind of

control and management that’s endemic to classical software, wherein data is
assumed to be passive and special-purpose objects (such as database manage-
ment systems) assume responsibility for validating all of the elements of data
they contain—with inordinate degrees of complexity as a byproduct.

Implementation
Having thought about all aspects of your objects, your stage, and your scripts
(the ones necessary to complete the story you are working on), you are ready
to write code—almost. If you are an XP developer, you are ready to write tests.
No matter what kind of developer you are, you need to think about two more
details: what your unit tests and code might look like and some additional infor-
mation that must be provided to those objects whose primary role is the main-
tenance of bits of knowledge (“data” objects).
 Chapter 9 All the World’s a Stage 287

Methods
For every message you listed on side 5 (message protocol) of the object cube,
you have to write a method and a test of that method (or tests). The first tests
are obvious: When the message is sent to the object, is the expected object
returned? Is it an instance of the right class? Is it in the correct state? By state, I
mean does it have a reasonable value; or, if it is a composite object, do all of its
instance variables have reasonable values?
The next set of tests concerns an object’s ability to recover from errors that
it might encounter between the time it’s asked to do something and the time it
returns an object in response to that request. These tests are trickier because
they usually involve a group of objects and a script. A very simple case would
be an error that occurs when an entryField object cannot validate itself with the
entered object. It must then invoke a dialog box object, ask the dialog to dis-
play an appropriate error message and accept an alternative value, and then
reperform the validation tests.
I have talked a lot about scripts and noted the possibility of script objects
actualizing the sequence of communications among a group of objects. In prac-
tice, a lot of the communications talked about in terms of scripts will be imple-
mented in methods. Thinking about scripts helps you design your next set of
tests. Does the precondition (trigger) implied in the script always cause the
appropriate message to be sent to the appropriate object? Does that object
implement the next bit of the script? Does that implementation satisfy the con-
ditions necessary to continue with the larger script?
The filling in of a form is an example of scripts being implemented in
methods rather than in an independent script object.

■ The form will receive the instantiate message. The form’s instantiate
message will be two lines (in some languages) of code: “Self, iterate
across all entry fields and tell them to instantiate themselves”; and
“Self, send yourself your validate message.”
■ Each entry field will receive the instantiate message. That method
will be about three lines: “Self, ask your source of input (an object in
one of your instance variables) to display itself and return a value to
you”; “Self, ask your validation rules to evaluate using the entered
value”; and “Self, let the form (whoever sent you the instantiate mes-
sage) know that you have successfully completed your job.”
288 Object Thinking

■ Each validation rule will receive the evaluate message. The evaluate
method will have two lines: “Self, instantiate” and “Self, resolve.”
■ Next the validation rule receives the instantiate message. Instantiate
consists of one line: “Self, tell all your elements that are variables to
instantiate themselves.”
■ Then the validation rule receives the resolve message. Resolve con-
sists of one line: “Send the message value to each of your terms in a
recursive fashion.”

Writing tests for each of these methods (script fragments) is relatively

straightforward—until you start to deal with exception and error handling. But
even this complication can be ameliorated by creating a new script that might
be implemented in several methods in other objects.
Your tests provide any additional specification you might need regarding
the form of your methods. Your selected (or mandated) programming language
provides you with the syntactic rules that must be adhered to in order to imple-
ment the semantics implied by your tests. Unfortunately, most programming
languages were not designed specifically to make it easy to express the seman-
tics that object thinking yields. You almost always will have to perform some
translation.

Knowledge Maintenance Objects

A lot (perhaps most) of the objects in your application will have the primary
charge of maintaining a bit of information. It’s convenient to call these data
objects—as long as you remember that they are not passive bits of data; they
are real objects!
In addition to the fact that it exists, each data object needs to know a lot
of additional information, information that traditionally was specified in a data
dictionary or a CRUD matrix.

Note A CRUD matrix gets it name from the actions create, read,
update, and destroy. It is a simple two-dimensional table that associ-
ates users and data items. Each data item in an application appears at
the head of a row of the matrix, and each user (user role) appears as a
column header in the matrix. Each intersecting cell then contains one
or more of the characters C, R, U, or D, reflecting that users have priv-
ileges (accesses) to the data item.
 Chapter 9 All the World’s a Stage 289

Items of importance from a data dictionary include

■ The name of each data item.

■ The composition of each data item—for example, Name is-composed-
of an optional Honorific, and a First Name, and an optional Middle
Initial, and a Last Name, and an optional Generational. (Of course,
the structure of a name will vary dramatically across cultures.)
■ The allowed size or upper and lower size limits; for example, a string
used for a First Name might have to contain at least 1 character but
no more than 35.
■ Any initial or default value.
■ A range of values—what data modelers call a domain; for instance,
a Zip Code data item would have a domain of all the values listed in
the United States Postal Service Zip Code Directory.

It will be essential to obtain this information about all your data objects.
But what do you do with it when obtained? Where is it implemented?
Earlier I suggested that certain classes called primitives are most likely to
be used to maintain knowledge—for example, characters, numbers, strings,
dates, times, and money. The obvious place to store CRUD and data dictionary
information is in the objects that use that information. However, it is unlikely
that you will actually want to modify all of the classes that can be used as prim-
itives. Instead, you might want to create a DataItem class.
A DataItem class is a way to reify passive data into a first-class object. An
instance of DataItem would have the following structure (Figure 9-21 shows the
object cube representation of this structure):
Label A string that names the data item.
Class The name of the class of which this data item is an instance.
Value The actual value of the data item. This would be an instance of one of
the primitive classes.
Composition An optional collection of DataItem objects that actualizes the
data dictionary definition.
Validation rules A collection of self-evaluating rules that would enforce
domain constraints.
Access rule A rule that enforces the line of the CRUD matrix pertinent to this
particular data item.
290 Object Thinking

Figure 9-21 Object cube representation of the DataItem class.

F09MQ21

If you elect the DataItem class option, you will then return to your object
cube, side 4, and replace a lot of the primitives listed there as DataItem objects.
Take care to replace only those primitives that are used more or less exclusively
as knowledge holders. If you are using a string or number, for instance, for
behavioral capabilities and not just to hold a value, you will not want to replace
them with DataItem objects.
 Chapter 9 All the World’s a Stage 291

Development at the Speed of Thought

Throughout this chapter (and the preceding two chapters), a lot of information,
models, heuristics, and details about objects have been presented. Every discus-
sion of every aspect of objects and object modeling included a caveat: every
step and every model is used only as a means of clarifying and communicating
understanding and is utilized only as part of the relevant XP activity.
Furthermore, there is an expectation that you will overtly use the material
presented in Chapters 7 through 9 only while learning object thinking. Eventu-
ally, all of the material in this entire book will be “inside your head,” and you
will be applying it without conscious awareness and while engaged in other
explicit activities—most notably while you are writing and refactoring code.
At first you will be consciously reminding yourself of the concepts and
ideas presented here and will make use of the various diagrams as explicit aids
to thinking. As you gain proficiency, you will use the aids less and less but still
might be seen “talking to yourself” just as a new reader will move his lips while
reading silently. Eventually you will just “do it.”
XP has three maturity levels—out of the box, adaptation, and transcen-
dence—that apply equally to object thinking. Once you have transcended the
material in this book, you will think faster and better about objects, and that
thinking will improve and expedite your thinking about design, testing, and
coding.
Wrapping Up
Despite any appearances to the contrary, objects are not something you do;
objects are a way that you think. To illustrate object thinking, it was necessary,
in preceding chapters, to distinguish between thoughts you would have at dif-
ferent points in the development process and to differentiate between thoughts
reflecting a holistic perspective and those that had a precise focus. It would be
easy, but wrong, to see phases, cookbook methods, and even prescriptive mod-
eling requirements in the last three chapters.
The material in Chapters 7 through 9 is an expression of the principles,
practices, values, metaphors, and the definitions presented in the first six chap-
ters. Object thinking requires developers to master the ideas so as to under-
stand the sample expressions and, more important, to be able, themselves, to
craft appropriate expressions of those ideas in response to the novel problems
they encounter in their own work.
An individual’s object thinking takes place inside her head—unobserved
by others, who can see only the product of that thought. The concerted object
thinking of a team is also mostly inside their heads, but it is sometimes manifest
in the communication products they produce and utilize as a kind of “external
group memory.” Ephemera like whiteboard drawings, story cards, poster
graphs, and visual metaphors and models are but byproducts of object thinking.
The product of object thinking is software that manifests simplicity and
composability, which lead, in turn, to adaptability, flexibility, and evolvability.
If object thinking is done in the context of XP (which contributes refactoring
and test-driven development techniques that complement object thinking) and

293
294 Object Thinking

agile development, the shared focus on the problem space and the needs of the
customer results in software that is usable and demonstrably supportive of busi-
ness/user requirements.
Unfortunately, there are circumstances in which it is difficult if not impos-
sible to apply object thinking.

Vexations
Sometimes you just can’t do the right thing. Or, more accurately, other consid-
erations make doing the right thing very hard. However diligent an object
thinker you become, you will encounter situations that tax your skills. The most
commonly encountered examples: the necessity of object applications using
relational databases for persistence and object-human communications via
graphical user interfaces.

The Impedance Mismatch Problem

Traditional computer thinking treats data as passive and helpless, manipulated
and moved by active procedures. Procedures, being a bit unruly, have been
known to abuse data, to make it something it shouldn’t be. Procedures also
tend to be egocentric and think that all data, its definitions and its values, are
properly determined by the procedure currently using that data. The result is
data that is inconsistent in definition, inconsistent in value, inaccessible because
some other procedure has possession, or needlessly duplicated.
The solution to these problems was to create a single repository for all
data—the database—and create one master procedure—the database manage-
ment system, or DBMS—to manage the data and protect it from “abuse” by
“unruly” application functions. Database philosophy is almost totally inconsistent
with the philosophy behind object thinking.
Databases violate the axiom that everything is an object. They violate the
principle that no object can do things to another object. Databases celebrate
centralized control. And relational databases, especially, are totally enamored of
the idea that everything can be formalized. It’s not surprising, therefore, to find
that objects and databases are frequently in conflict with each other.
Some examples of the kinds of problems encountered when mixing
objects and relational databases are as follows. (Extended relational and object
databases have been modified and extended in their capability to ameliorate
most of the problems noted.)
 Chapter 10 Wrapping Up 295

■ Objects are not defined based on their attributes, but entities and the
tables that they inhabit in a database are defined in precisely that
manner. Objects can store, in a variable, objects whose primary pur-
pose is to provide a calculation or a service rather than to embody a
piece of data. The rules that you follow to make something “persist-
able” vary radically between the object world and the database
world.
■ Objects have variables that can contain arbitrarily complex objects,
including collections and multivalued constructs. Databases can
store only a defined, and limited, set of primitive types. For an object
to be stored in a database, it must be disassembled and translated,
making both storage and retrieval complicated and error-prone pro-
cesses. Several companies have made significant amounts of money
providing special-purpose “object-relational mapping” software that
“solves” this disassembly-and-translation problem.
■ If a running application uses objects that persist in a database, there
is no assurance that an object’s value in the application is consistent
with its value in the database. Other applications might be updating
the database, and there is no way for the database to notify the
object that its persistent value has changed through intervention by
another program. (Extended relational databases now have the abil-
ity to notify objects of this kind of event.)
■ Databases do not preserve the encapsulation of an object. While an
object is persistent (in the database), all of its contents, the objects
stored in the object’s variables and in some cases even the object’s
methods, are available for direct access by any other program using
that database. This leads to all kinds of potential coupling between
objects (based on program-specific contexts), which interferes with
(destroys) the modularization arising from good object design in the
first place.

In an ideal object world, there would be no need for databases. Every

object would have its own URL.1 Persistent storage devices would offer nothing
more than space and a path extension for the object’s URL so that it could be
located when necessary. The object would serialize itself (marshal itself into a

1. URL—Uniform Resource Locator—is a technical term for the address of resources on the Web. I am
using it more colloquially, still meaning a unique address with the implication that such an address
would be as simple as a URL and be independent of physical implementation (i.e., not rely on
platform-specific offsets and extents).
296 Object Thinking

bit stream) and de-serialize itself in order to reside on the storage medium.2 The
object would also register its location with a “white pages” object so others
could find it when necessary. (The white pages would function much as the
domain name servers employed to match text and numeric URLs for the Web—
that is, they would be hierarchical and replicated for efficiency of access.)

Why Databases?
Persistence is not the only reason for using a database. Databases also
offer computational advantages for certain application needs. For exam-
ple, they provide functions for efficiently and quickly tabulating and com-
puting across large collections of instances (tuples, or rows, in database
vernacular). If I had a large collection of customer objects (a few million,
say) and wanted to know which of them lived in a particular postal code,
I would have to ask each object in turn where it lived and have it report
to an output collection if it lived in the specified postal code. This would
take a long time. If that same collection were stored as tuples in a rela-
tional database, a simple vector operation would extract the relevant
objects in short order. On the other hand, storing a complex object—say
the bill of materials for a passenger jet, perhaps—in a set of relations in a
database and then asking that database to print a list of all components
used to build that airplane would take a long time for the database but
would be a relatively trivial operation for the object. In the main text of
this section, concern is focused on persistence. But object thinking design-
ers need to give attention to the functional advantages of databases as well
as their persistence services.

Services that a database is supposed to provide to data—access control

and enforcement of integrity constraints, for instance—would be assumed by
each object. An object could queue messages or simply indicate a busy signal if
it received multiple messages simultaneously. Each object could carry within
itself its own set of rules (self-evaluating rule objects) that would ensure the
integrity of its value and allow it to reject messages from sources that have not
been given access authorization. (Think in terms of access tables maintained by

2. The object-relational mapping software mentioned in this section provides a kind of marshalling
where the serialization results in SQL or proprietary data manipulation statements rather than a bit
stream suitable for writing to magnetic media.
 Chapter 10 Wrapping Up 297

most operating systems that allow some users access to files while denying it to
others. The entries in such tables are merely transferred to the objects instead of
being maintained in a central place.)
Until the object vision is realized in full, it will be necessary to compro-
mise and accommodate and make use of existing database systems. Treating
the database as a great big object—employing the concept of putting an object
wrapper around a nonobject thing—is one way to effect that compromise.

A Problem with GUIs

As noted previously (sidebar, Chapter 8), the advent of graphical user interfaces
promoted object thinking while simultaneously distorting what it meant to be
an object. The primary problem was trying to design object applications from
the interface in. The first step advocated by many object enthusiasts was to
sketch or prototype the application interface and then figure out what kind of
objects were needed behind the scene. This kind of thinking also reflects pre-
GUI computer thinking, wherein it was common to design the (character-
based) forms and reports before writing the application code that would use
them for input and output.
Object thinking says that a view (sometimes called an interface)3—graphical
or otherwise—is a means for an object to communicate with another object and
nothing more. The need for a view arises when an object needs to present itself
in a “non-native” form to some other object (usually a human being) or applica-
tion (for example, an XML view for data objects being shared across platforms).
Discovery of the need and the parameters that must be satisfied by a view
is manifest in the scenarios in which the object participates. Whenever an object
is asked to display itself, it must use a view—a representation—appropriate for
the sender of that display message. If, for example, an object is trying to instan-
tiate itself (get a value for itself), it must present a view of itself as an implicit
request to a human being (or other service-providing object) for a value. If we
are building a GUI that will serve as an intermediary between a software object
and a human object, we will use glyphs for display purposes and widgets for
interaction purposes.
But which glyphs and widgets need to be included in the GUI? Only those
necessary to complete the scenario or scenarios4 of immediate interest as the

3. The term interface is used, unfortunately, in more than one way. It refers to the object-specific mes-
sage protocol plus event notification information as recorded on sides 5 and 6 of an object cube. It
also means the form assumed by an object in order to present itself to a human user or another appli-
cation object. We are using the second sense of the word in this section.
4. In most applications, several objects will be simultaneously presenting themselves to or requesting
interaction with the human user. The totality of the GUI will therefore be composed of multiple object
views. This was noted in the discussion of the MVC architecture in Chapter 9.
298 Object Thinking

application runs. This perspective is counterintuitive for most developers

because it suggests that a GUI be defined from the application out.
As an example, consider a brewery. Off to one side are vats filled with
beer. At the other side are trucks waiting to take cases of brew to avid custom-
ers. In between is a complex production line consisting of bottle washers, filler
stations, capping machines, and package assemblers. Above it all is a control
station that monitors the brewery and notifies human managers of status and
problems. Traditional developers are likely to begin their analysis and design of
“a brewery management system” from the point of view of the control panel.
This is analogous to designing from the interface in.
Object thinking would suggest, instead, that you consider which object is
the prime customer of the brewery and all its myriad machines. On whose
behalf does the complex maze of equipment exist? The correct business answer
is, of course, “The customer.” But an answer more reflective of object thinking
is, “The beer.” All scenarios are written from the perspective of the beer, trying
to get itself into a bottle, with a cap, placed in a package, and resident in a
truck. The control panel is a passive observer5 of the state of the brewery. If the
beer encounters a problem at some point, it’s the responsibility of the beer to
request intervention of the human operators by sending a message to the control
panel (or machine-specific control panels) requesting an intervention service.
This perspective will simplify GUI design and, more important, eliminate
the host of manager and controller objects that seem to inevitably arise when
designing from the control panel’s (GUI’s) perspective.
Other vexations for object thinkers include programming language imple-
mentations that do not allow direct expression of object concepts or that
impose safety and control structures that are antithetical to object thinking.
Operating systems and hardware also present situations forcing compromise of
object principles in order to get anything done. Dealing with these vexations in
the most efficacious manner possible requires application of object philosophy,
metaphor, and principles.
Dealing with negativity is no way to end a book on object thinking, so we
will redirect our attention to some issues that extend the ideas presented in ear-
lier chapters.

5. In an experimental microbrewery, the panel (in conjunction with the brewmeister) might be more
than a passive observer. It might also be a source of instruction to the beer to modify its behavior. It
is still not a controller—merely another interested player that provides services to the beer in its effort
to “be the best brew it can be.”
 Chapter 10 Wrapping Up 299

Extensions
One of the tenets of extreme programming involves the avoidance of prema-
ture abstraction. Abstraction, as they are using the term, is a kind of optimiza-
tion that arises from refactoring code. It’s not necessary, however, for every
abstraction to be rediscovered every time it’s used. Abstractions, once discov-
ered, can be remembered and applied in similar circumstances in the future.
This has the effect of accelerating thought and thereby accelerating develop-
ment—something quite consistent with XP principles and values.
For example, the stereotypes discussed as part of side 2 (the description)
of an object cube are abstractions. If you are developing a financial system and
encounter an object named portfolio, your knowledge of collections—an
abstraction—should lead you to discover that a portfolio is just a collection with
some added behavior. Patterns and architectures are another form of abstrac-
tion that you can learn, as abstractions, and then apply as you think about
design and implementation.6 In Chapter 9, some architectural abstractions were
introduced (for example, MVC, blackboards, pipes, and filters). Patterns have
also been introduced—for example, the discussion of blind dispatching, which
is an example of the observer or publish-and-subscribe pattern—but a full
exploration of patterns is best pursued in the patterns literature itself. (See the
bibliography for starting points.)

Frameworks
A framework is another kind of abstraction, somewhere between a design pat-
tern and a full architectural pattern. Frameworks are organizational abstractions
that provide a set of objects with expected behaviors and the relationships that
enable them to interact in order to achieve some collective goal. Good frame-
works provide the essential core of a programming solution. Extension of the
framework by adding other objects and other relationships is expected. How-
ever, a good heuristic is to demand unassailable justification before accepting
any addition to the framework. See if the unaltered framework can solve your
problem before you embellish it. Following are some examples of frameworks.

6. See the work of Alan Shaloway and of Josh Kerievski on refactoring to patterns and Bob Martin’s Agile
Software Development for discussion of how and why patterns serve as guides and targets, not as tem-
plates that get dropped into your design. Blind composition of patterns actually leads to very poor
code.
300 Object Thinking

Composable Document
A composable document has variable content and constraints on the appear-
ance of the content, which are sometimes determined by the presence or
absence of other content. A report is a kind of composable document. A legal
form, like a loan disclosure document, is another.
The composable document framework (Figure 10-1) suggests that a com-
posableDocument is a collection of elements (which means it has the behavior
of adding, deleting, and providing access to the collected elements) that can
identify itself, describe itself, and display itself. Display requires collaboration
with elements, which can display themselves. Both the document and its ele-
ments have extent (the amount of space occupied by the element—for a docu-
ment, usually the two-dimensional area), which can be reported and, in the
case of the document, allocated to an element.

ID Has Has Extent

Document
Has
Description
Contains Has

Constant Has Element

take-the-
form-of-a Variable
Has
Has
Rule Has Constant
Has
Operator

Figure 10-1 Composable document framework.

F10MQ01

Both elements and the document must satisfy certain composition rules,
which are embodied in self-evaluating rules. (The composition of a rule as a
collection of variables, constants, and operators is shown in Figure 10-1.) This
framework provides all of the objects necessary to model the construction of
any form or report. (Actual display requires a host of additional objects, glyphs,
bitmaps, vector graphics, and so on—a vexation in that way too much work is
required to paint a screen than would be the case if object thinking had guided
the developers of operating systems and graphics libraries.)
 Chapter 10 Wrapping Up 301

Object Routing and Tracking

The object routing and tracking framework (Figure 10-2) is useful as an abstrac-
tion for the core description of systems as disparate as a network packet man-
ager and a UPS package delivery system.

Network

Address
a-collection-of Link
has
Origin has
Node has Characteristics
Destination
is-ordered- is-an-ordered-collection-of
by
must match/
Route be compatible with

uses/requires

has
has Object has Description

Figure 10-2 Object routing and tracking framework.

F10MQ02

This framework suggests that an object has an originLocation, a destina-

tionLocation, a description, and a unique ID. It obtains a route from a network
(an object that specializes in solving np-complete (traveling salesman type)
problems. A route is simply an ordered set of nodes and links that connect an
originLocation to a destinationLocation. Nodes and links have addresses and
characteristics. An object obtains a route, checks with both nodes and links to
track its own progress, and raises an alert if it detects a deviation from its
assigned route. This alert enables making a request for an alternate route. Var-
ious parameters, which match object description with link or node characteris-
tics, time or priority constraints, and so forth, are encapsulated in rules
associated with the appropriate objects.

Resource Allocation and Scheduling

Resource allocation is important in numerous business systems. Scheduling is
fundamentally nothing more than a kind of resource allocation, with the
resource being time. Three objects form the core of this framework (shown in
Figure 10-3):
302 Object Thinking

■ Resource, which can have other resources (such as a calendar with

fine-grained time divisions) as part of its structure and which has a
description and an identity
■ Request, which can recursively be a collection of requests and which
also has a description (resource requirements), an identity, and a
Requester
■ Reservation, which associates a resource with a request and has an
identity along with references to the objects it is associating

Reservation

Has
Request Resource

Has Has

Requirements Must Description

Match

Figure 10-3 Resource allocation and scheduling framework.

F10MQ03

Numerous other frameworks can be constructed, each of which captures

and embodies the best of object thinking. These abstractions can then be used
as shortcuts in future efforts to apply object thinking to solve new problems.
A very different kind of extension to object thinking arises from the prin-
ciple that objects are simulations. An object does not represent, in some abstract
way, a thing in the real world; it simulates that real-world thing. Objects are
evocative rather than representational. This is why Alan Kay (among others)
speaks of the need for objects to preserve the “user illusion” and behave and
appear in a manner consistent with the thing being simulated. This evocative
property of objects can be utilized in interesting ways—including the depiction
of system architecture.

Object-Based Evocative Architecture

Figure 10-4 is a photo of a Thangka painting on the wall of my office. This
painting is an evocative model of Tibetan Buddhist cosmology and philosophy.
Even a cursory glance reveals that it has some kind of structural organization, a
significant amount of detail, and lots of interesting and unusual images. Less
obvious is that it is a very good object model. The images on the painting are
 Chapter 10 Wrapping Up 303

representational in part. Each icon and each organizational segment of the

whole can be seen as representing a particular deity or circumstance, but that is
almost coincidental—a case of convenient labeling. The real purpose of the
painting as a whole, and of each individual element of the painting, is to evoke
memories, primarily memories of stories the viewer of the painting was told as
he or she grew up in a Tibetan household. As complex as the painting itself
appears to be, that complexity pales in contrast with the volume of memories
and intertwined stories that the painting recalls in the mind of the viewer.

Figure 10-4 Thangka painting.

F10MQ04.gif
304 Object Thinking

Given that XP eschews the development of formal representational archi-

tectures as a prelude to development and that some kind of holistic view of the
system to be built is desirable, perhaps an evocative model patterned after the
Thangka painting and reflecting the simulation aspect of objects will provide a
useful informal tool.
Essential elements of the painting that would have equivalents in the XP
architectural model include

■ A center circle with icons recalling to mind the driving forces behind
the system under development. In the painting, these are icons
reminding us of attachment, greed, anger, and so on. In a system
model, these would remind us of the importance of money (almost
inevitable), a particular client (a CEO perhaps) who will pass final
judgment on our work, a customer, or a service.
■ A large segmented circle, each segment representing some kind of
portioning of the system, preferably isomorphic to some degree with
the actual portioning of the business for which the system is being
built. In the painting, these are the various realms of existence—for
example, heaven, hell, material world. In our model, these might
represent realms such as order entry, inventory, accounting, and
manufacturing; or segments of a smaller-scale system—data entry,
back-end processing, backup and recovery, and so forth.
■ Icons of various sorts arranged in a tableau evocative of the stories
that relate those icons to one another. Each icon evokes a specific
story or set of stories about a particular element of the overall sys-
tem. Icons can appear in more than one segment.
■ A narrower outer circle, also segmented, with each segment repre-
senting a stage in the “circle of life.” This circle, in the painting, is the
progress from birth to death that each of us traverses. In a system
architecture, each segment would represent a stage in the processing
of whatever the system is intended to support or perhaps some kind
of business cycle from prospect to customer or raw material to prod-
uct.
■ Finally, outside the circle, icons that recall stories about outside influ-
ences, things or forces or people that can affect our system but are
outside the scope of what we can actually build.

Dynamism can be added to the model in many different ways. If each icon
evokes a particular story and the work required to realize that story, the orienta-
tion of the icon (down = not started, right angle left = in progress, vertical = done,
 Chapter 10 Wrapping Up 305

right angle right = abandoned dead end) can be used to depict status. A quick
glance at the model reveals a rough approximation of total effort and total
achievements. This is the same information that appears on many an XP
project’s walls and whiteboards but in a more compact form.
The dramatic form and the unusual context behind the presentation of this
architecture might obscure the fact that it’s merely a pictorial representation of
the way that object thinkers think of objects and applications: as stories, à la XP.

Provocation—The Future of Objects

Thinking can be focused toward the accomplishment of a task. All of the mate-
rial in this book so far is intended to aid developers in that kind of thought. To
conclude, however, I would like to provoke a different kind of thinking—spec-
ulative thinking. Specifically, what might happen if all of the ideas, philoso-
phies, and metaphors presented in this book were taken to the extreme? The
following story is one possibility.
Circa 1998 a small group of programmers decided they were “mad as hell
and not going to take it anymore.” Although the rebellion came in many flavors,
the most prominent example of it was extreme programming (XP). Similar, con-
currently active groups considered themselves agile and formed, with the
XPers, the Agile Alliance. Xgilista is an umbrella term covering the spectrum of
extreme and agile approaches to software creation.
Among other things, Xgilistas rejected the prevailing idea that software
projects, particularly large mission-critical projects, required massive amounts
of up-front analysis, design, and specification. They claimed that it was quite
possible to realize such systems with an incremental process that was highly
disciplined, human-centric, social, and ethical.
Large, successful systems emerged from incremental development—exist-
ence proofs of yet another old idea. Grady Booch, in Object-Oriented Design,
1991, stated that everyone knew that the only way to build successful large sys-
tems was to evolve them from successful small ones.
In rejecting up-front design, Xgilistas essentially rejected the need for
large, monolithic programs. They challenged, implicitly and explicitly, the
assumption that large systems were actually required.
XP focus on delivering the simplest possible deliverable capable of meet-
ing a real but discrete business need had an unintended side effect. (At least,
this result was never explicitly stated as an expected outcome of XP, but one
should not underestimate the sly, subtle, and covert intent of a Ward Cunning-
ham or a Kent Beck.) Users started to see their requirements in a very different
light. Instead of automatically thinking they needed huge, monolithic software
306 Object Thinking

systems, they started to request a set of small, targeted software-based business

tools instead. The perceived role of software changed dramatically. Users
started to ask for objects capable of providing specific and limited services. The
utility of such objects was based on how limited an intrusion they made into
the business process and how faithfully they simulated a business role in that
process.
Objects (applications are just composite objects with collective responsi-
bilities) came to be seen as analogous to new employees. Just as you would
hire a human being with specific skills to perform some well-defined job func-
tion, so too businesses started to “hire” individual objects instead of massive
systems.
In retrospect, this outcome should also have been obvious. In the late
1800s, Pareto’s 80/20 “law” was formulated. Empirical study of the actual use of
programs such as Microsoft Windows, Microsoft Word, WordPerfect, and so on
revealed a variation of Pareto’s Law—80 percent of the users used 20 percent or
less of the functionality offered by such programs. No one used all of the fea-
tures. So why were they built? Why did companies spend millions of dollars on
monstrous dinosaurs such as SAP and its kindred “integrated applications”? Mar-
ket forces and management folly is the short answer.
Until Xgilistas demonstrated differently, advocates of monolithic formal
systems claimed that the incremental, bottom-up approach could not yield, at
least efficiently, systems that were capable of controlling large businesses or
complex systems such as air traffic control. The XP mantra, “Architecture
emerges,” was dismissed as absurd. But architecture did emerge. Executives
and even entire corporations are, after all, just another kind of user with
focused business needs. Satisfaction of those needs did require aggregation and
distillation of results generated by other objects and job processes. But this was
addressed with judicious use of the observer pattern.
Xgilistas finally broke the back of the prevailing software engineering and
computer science myth—to wit, software is formally defined and constructed.
Users, and even managers, were finally ready to accept what objects had
offered 30 years earlier.
Enter the Objectionary, a combination of dictionary and object factory.
The Objectionary is nothing more than the product of taking object ideas—as
originally voiced—seriously. The salient features of the Objectionary include
the following:

■ The Objectionary itself is a kind of “object vending machine,” dis-

pensing objects of various types upon request.
 Chapter 10 Wrapping Up 307

■ The total number of Objectionary objects is less than 2000, a number

that continues to be sufficient to assemble every software application
attempted. (Remember the Lego Brick metaphor and the discussion
of how many objects—classes, really—are necessary.)
■ Each object is an autonomous executable entity. Smalltalk objects are
freed from their dependence on image environments.
■ Every object has a unique ID and a unique “address”—a realization
of Alan Kay’s suggestion that every object have its own URL.
■ Objects are nothing more than collections of objects. After all, every-
thing is an object. This means, of course, that messages and methods
become first-class objects in their own right.
■ Objects are fully executable (“precompiled byte code strings”) but do
require a hardware-specific Virtual Machine (VM) for access to
primitive hardware services.
■ There are a few key abstractions, such as the self-evaluating rules
conceived of by West and implemented by Johnson. Their work
essentially reified source code statements into first-class objects that
could be modified at execution time.

Access to the Objectionary almost eliminated one profession—that of pro-

grammer—and replaced it with another, application assembly specialist. (Again,
remember the Lego Brick metaphor and its differentiation between creators and
assemblers.) These specialists adopted the rather ugly, but descriptive, appella-
tion of collagist, reflective of the kind of artist who creates new works by
assembling materials.
End user computing—for real, not the pseudoprogramming that had sti-
fled earlier forays into this area—became all the rage. User-created software
was quite analogous to the Lego Brick constructions of children—workable but
lacking in certain qualities. Some novice assemblers had a knack for putting
together objects in such a way that the resultant assemblies not only were func-
tional but were starting to become beautiful. Others noticed and were more
than willing to compensate such “hobbyists” or “tinkers” to assemble on their
behalf.
Software assembly became a craft, just as programming once had been.
Collagists formed a community of shared expertise. But this time the art went in
a much different direction than programming had almost 50 years before. The
expertise that set a collagist apart from other software assemblers centered on
308 Object Thinking

composition, proportion (e.g., the Golden Mean), aesthetics, efficiency (not the
rabid concern with machine efficiency of C and C++ programmers—rather, the
efficiency of interaction among humans and objects, and between objects and
objects), and simplicity or elegance. Not being amenable to quantification and
hence formalist corruption, these skills flourished as they had in other artistic
endeavors for centuries.
The basic tools of collagism were all open source, free to anyone to use
and modify. Collagists made money doing what “anyone could do for free” sim-
ply because they had “artistic skill and style.” People were more than willing to
pay for style, just as they did with regard to the other arts. It was a matter of
prestige to have an original “Jeffries payroll system,” and companies were will-
ing to pay for that prestige, just as they did for the Mattisse hanging in the
lobby.
Two markets have come to thrive, side by side: Generic “house-brand”
assemblies and designer assemblies. Users bought what they could afford and
what they most needed. In many cases, generic was more than adequate, but
for some highly visible (or core competency) applications, designer quality was
required—and paid for.
Aspiring collagists shunned software education offered by traditional
departments and universities (even when they stopped calling themselves soft-
ware engineering programs) in favor of apprenticeship programs located in lib-
eral arts colleges.
Objectionary objects turned out to be self-modifiable. A key idea—self-
evaluating rules—was the only prerequisite. Rules can be recursively defined
and can be modified simply by adding, deleting, or reordering the member
objects. Self-evaluating rules are well suited for use as genetic algorithms.
Shortly after the release of the Objectionary, various assembly hackers noticed
this fact and took advantage of it to create objects that reconfigured themselves
as a function of the environment in which they operated.
The earliest self-modifying software consisted of nothing more than com-
plex rules that evaluated to some result that could be subjected to a fitness test.
Later attempts allowed participation of all sorts of objects, applications (as it
were) that were free to modify the collection of participating objects, the script
object that coordinated the interaction of the objects in that collection, and the
rules contained within each object that constrained or enhanced its individual
behavior.
An interesting story, or a plan of action for Object Thinkers and Xgilistas?
Bibliography
(Books marked with an * are particularly relevant to the philosophical underpinnings of
object thinking as discussed in the first chapters of this book.)

Albin, Stephen T. The Art of Software Architecture: Design Methods and Techniques.
Indianapolis, IN: Wiley Publishing, Inc., 2003.
*Alexander, Christopher. Notes on the Synthesis of Form. Cambridge, MA: Harvard University
Press, 1964.
Alexander, Christopher, Sara Ishikawa, and Murray Silverstein. A Pattern Language: Towns,
Buildings, Construction. New York: Oxford University Press, 1977.
*Alexander, Christopher. The Timeless Way of Building. New York: Oxford University Press,
1979.
*Alexander, Christopher. The Nature of Order: Book One, The Phenomenon of Life. New
York: Oxford University Press, 2001.
Ambler, Scott W. The Object Primer: The Application Developer’s Guide to Object-Orientation.
New York: SIGS Books, 1995.
Anderson, Jim, George Bosworth, Alberto A. Della Ripa, Barbara Noparstak, and Michael
Teng. Methods Owners Manual. Digitalk,1985.
Arranga, Edmund C., and Frank P. Coyle. Object Oriented COBOL. New York: SIGS Books,
1996.
Arthur, Lowell Jay. Rapid Evolutionary Development: Requirements, Prototyping & Software
Creation. New York: John Wiley & Sons, 1992.
Bahar, Mory. Object Technology Made Simple. East Greenwich, RI: Simple Software
Publishing, 1996.
Bapat, Subodh. Object-Oriented Networks: Models for Architecture, Operations, and
Management. Englewood Cliffs, NJ: PTR Prentice Hall, 1994.
Bassett, Paul G. Framing Software Reuse: Lessons from the Real World. Upper Saddle River,
NJ: Prentice Hall PTR, 1997.
Baudoin, Claude, and Glenn Hollowell. Realizing the Object-Oriented Lifecycle. Upper Saddle
River, NJ: Prentice Hall PTR, 1996.

309
310 Object Thinking

Beaudouin-Lafon, Michel. Jack Howlett, translator. Object Oriented Languages: Basic

Principles and Programming Techniques. London: Chapman and Hall, 1994.
Beck, Kent. Guide to Better Smalltalk: A Sorted Collection. SIGS Reference Library. Cam-
bridge: Cambridge University Press, 1999.
*Beck, Kent. Extreme Programming Explained: Embrace Change. Boston: Addison-Wesley,
2000.
Beck, Kent, and Martin Fowler. Planning Extreme Programming. Boston: Addison-Wesley,
2001.
Bellin, David, and Susan Suchman Simone. The CRC Card Book. Reading, MA: Addison-
Wesley, 1997.
Berard, Edward V. Essays on Object-Oriented Software Engineering. Vol 1. Englewood
Cliffs, NJ: Prentice Hall PTR, 1993.
Bergin, Joseph. Data Abstraction: The Object-Oriented Approach Using C++. New York:
McGraw-Hill, Inc., 1994.
Blaha, Michael, and William Premerlani. Object-Oriented Modeling and Design for Database
Applications. Upper Saddle River, NJ: Prentice Hall, Inc., 1998.
*Bohm, David. Wholeness and the Implicate Order. London: Routledge and Kegan Paul,
1980.
Borland. The World of C++: The Fastest Way to Become a ++ Programmer. Borland
International, 1991.
Bourne, John R. Object-Oriented Engineering. Building Engineering Systems Using
Smalltalk-80. Boston: Irwin, 1992.
Bowman, Charles F., ed. Wisdom of the Gurus: A Vision for Object Technology. New York:
SIGS Books, 1996.
*Brown, John Seely, and Paul Duguid. The Social Life of Information. Boston, MA: Harvard
Business School Press, 2000.
Brown, William J., Raphael C. Malveau, Hays W. “Skip” McCormick III, and Thomas J.
Mowbray. Anti-Patterns: Refactoring Software, Architectures, and Projects in
Crisis. New York: John Wiley & Sons, Inc., 1998.
Budd, Timothy. An Introduction to Object-Oriented Programming. Reading, MA: Addison-
Wesley Publishing Co., 1991.
Buhr, R. J. A., and R. S. Casselman. Use Case Maps for Object-Oriented Systems. Upper Saddle
River, NJ: Prentice Hall PTR, 1996.
*Bulhof, Ilse N. Wilhelm Dilthey: A Hermeneutic Approach to the Study of History and Culture.
The Hague: Martinus Hijhoff, 1980.
 Bibliography 311

Burleson, Donald K. Practical Application of Object-Oriented Techniques to Relational

Databases. New York: John Wiley and Sons, Inc., 1994.
Burnett, Margaret M., Adele Goldberg, and Ted G. Lewis, eds. Visual Object-Oriented Pro-
gramming: Concepts and Environments. Greenwich, CT: Manning Publications
Co., 1995.
Cattell, R. G. G. Object Data Management. Reading, MA: Addison-Wesley Publishing
Company, 1991.
Chorafas, Dimitris N., and Heinrich Steinmann. Object-Oriented Databases. Englewood
Cliffs, NJ: Prentice Hall PTR, 1993.
Coad, Peter, and Ed Yourdon. Object-Oriented Analysis. Englewood Cliffs, NJ: Prentice-
Hall, 1991.
Coad, Peter, and Ed Yourdon. Object-Oriented Design. Englewood Cliffs, NJ: Prentice-Hall,
1991.
Coad, Peter, with David North and Mark Mayfield. Object Models: Strategies, Patterns and
Applications. Englewood Cliffs, NJ: Prentice Hall, 1995.
Coad, Peter, Eric Lefebvre, and Jeff DeLuca. Color Java Modeling with UML: Enterprise
Components and Process. Englewood Cliffs, NJ: Prentice-Hall, 1999.
Cockburn, Alistair. Writing Effective Use Cases. Boston: Addison-Wesley, 2001.
*Cockburn, Alistair. Agile Software Development. Boston: Addison-Wesley, 2002.
Coleman, Derek, Patrick Arnold, Stephanie Bodoff, Chris Dollin, Helena Gilchrist, Fiona
Hayes, and Paul Jeremaes. Object Oriented Development: The Fusion Method.
Englewood Cliff, NJ: Prentice Hall, 1993.
Collins, Dave. Designing Object-Oriented User Interfaces. Redwood City, CA: Benjamin
Cummings Publishing Co., Inc., 1995.
Cook, Steve, and John Daniels. Designing Object Systems: Object-Oriented Modeling with
Syntropy. New York: Prentice-Hall, 1994.
Coplein, James O., and Douglas C. Schmidt, eds. Pattern Languages of Program Design.
Reading, MA: Addison-Wesley Publishing Co., 1995.
Cox, Brad. Object-Oriented Programming: An Evolutionary Approach. Reading, MA:
Addison-Wesley Publishing Company, 1986.
Cox, Brad. Superdistribution: Objects as Property on the Electronic Frontier. Reading, MA:
Addison-Wesley Publishing Co., 1996.
*Coyne, Richard. Designing Information Technology in the Postmodern Age: From Method
to Metaphor. Cambridge, MA: MIT Press, 1995.
312 Object Thinking

*David, Alan M. 201 Principles of Software Development. New York: McGraw-Hill, Inc.,
1995.
*Dahlbom, B., and L. Mathiassen, Computers in Context: The Philosophy and Practice of
Systems Design. Oxford: Blackwell, 1993.
De Champeaux, Dennis, Douglas Lea, and Penelope Faure. Object-Oriented System
Development. Reading, MA: Addison-Wesley Publishing Company, 1993.
Deitel, H. M., P. J. Deitel, J. A. Listfield, T. R. Nieto, C. H. Yaeger, and M. Zlatinka. C#: A
Programmer’s Introduction. Upper Saddle River, NJ: Prentice-Hall, 2003.
Digitalk, Inc. Smalltalk/V. Los Angeles, CA: Digitalk, Inc., 1986.
*Dittrich, Yvonne, Christiane Floyd, and Ralf Klischewski. Social Thinking—Software
Practice. Cambridge, MA: MIT Press, 2002.
Dorfman, Len. Object-Oriented Assembly Language. Blue Ridge Summit, PA: Windcrest,
1990.
*Dreyfus, Hubert L. Husserl, Intentionality, and Cognitive Science. Cambridge, MA: MIT
Press, 1982.
*Dreyfus, Hubert L., and Stuart E. Dreyfus with Tom Athanasiou. Mind over Machine. New
York: The Free Press, a division of Macmillan, Inc., 1985.
Eeles, Peter, and Oliver Sims. Building Business Objects. New York: John Wiley and Sons,
Inc., 1998.
*Ehn, Pelle. Work-Oriented Design of Computer Artifacts. Stockholm: Arbetslivcentrum,
1988.
Ellis, John R. Objectifying Real-Time Systems. New York: SIGS Books, 1994.
Embley, David W., Barry D. Kurtz, and Scott N. Woodfield. Object-Oriented Systems
Analysis: A Model Driven Approach. Englewood Cliffs, NJ: Prentice Hall PTR,
1992.
Evans, Eric. Domain-Driven Design: Tackling Complexity in the Heart of Software. Boston:
Addison-Wesley, 2004.
*Feyerabend, Paul. Against Method. London: Verso, 1975.
Fingar, Peter. The Blueprint for Business Objects. New York: SIGS Books, 1996.
Firesmith, Donald G. Object-Oriented Requirements Analysis and Logical Design: A
Software Engineering Approach. New York: John Wiley and Sons, Inc., 1993.
*Floyd, C., H. Zullighoven, R. Budde, and R. Keil-Slawik. Software Development and Reality
Construction. Berlin: Springer-Verlag, 1991.
 Bibliography 313

Fowler, Martin. Analysis Patterns: Reusable Object Models. Menlo Park, CA: Addison-Wesley
Longman, Inc., 1997.
Fowler, Martin. Refactoring: Improving the Design of Existing Code. Reading, MA: Addison-
Wesley, 1999.
Frolund, Svend. Coordinating Distributed Objects: An Actor-based Approach to Synchroni-
zation. Cambridge, MA: MIT Press, 1996.
*Gabriel, Richard P. Patterns of Software: Tales from the Software Community. Oxford:
Oxford University Press, 1996.
*Gadamer, Hans-George. Philosophical Hermeneutics. Translated and edited by David E.
Linge. Berkeley: University of California Press, 1976.
*Gadamer, Hans-George. Reason in the Age of Science. Translated by Frederick G.
Lawrence. Cambridge, MA: MIT Press, 1982.
Gale, Thornton, and James Eldred. Getting Results with the Object-Oriented Enterprise
Model. New York: SIGS Books, 1996.
Gamma, Erich, Richard Helm, Ralph Johnson, and John Vlissides. Design Patterns: Elements
of Reusable Object-Oriented Software. Reading, MA: Addison-Wesley, 1995.
*Glass, Robert L. Software Creativity. Englewood Cliffs, NJ: Prentice-Hall PTR, 1995.
*Glass, Robert L. Facts and Fallacies of Software Engineering. Boston: Addison-Wesley,
2003.
Goldberg, Adele, and Ken Rubin. Succeeding with Objects: Decision Frameworks for Project
Management. Reading, MA: Addison-Wesley Publishing Co., 1995.
Gossain, Sanjiv. Object Modeling and Design Strategies. SIGS Books. Cambridge: Cambridge
University Press, 1998.
Graham, Ian. Migrating to Object Technology. Wokingham, England: Addison-Wesley
Publishing Co., 1994.
Grand, Mark. Patterns in Java, vol 1. New York: John Wiley and Sons, Inc., 1998.
Greenbaum, Joan, and Morten Kyng. Design at Work: Cooperative Design of Computer
Systems. Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers, 1991.
Henderson-Sellers, Brian, and Julian Edwards. The Working Object. Sydney, Australia: Prentice-
Hall, 1994.
Henderson-Sellers, Brian. Object-Oriented Metrics: Measures of Complexity. Upper Saddle
River, NJ: Prentice Hall PTR, 1996.
Highsmith III, James A. Adaptive Software Development: A Collaborative Approach to
Managing Complex Systems. New York: Dorsett House Publishing, 2000.
314 Object Thinking

Highsmith, Jim. Agile Software Development Ecosystems. Reading, MA: Addison-Wesley,

2002.
Hill, David R. C. Object-Oriented Analysis and Simulation. Harlow, England: Addison-
Wesley Press, 1996.
Hiltzik, Michael A. Dealers of Lightning: Xerox Parc and the Dawn of the Computer Age.
New York: Harper Collins, 1999.
Holmes, Jim. Object-Oriented Compiler Construction. Englewood Cliffs, NJ: Prentice Hall
Inc., 1995.
Hughes, John G. Object-Oriented Databases. New York: Prentice-Hall, 1991.
IBM. Object Oriented Interface Design. New York: Que Books, 1992.
Islam, Nayeem. Distributed Objects: Methodologies for Customizing Systems Software. Los
Alamitos, CA: IEEE Computer Society Press, 1996.
Jacobson, Ivar. The Object Advantage: Business Process Reengineering with Object Technology.
ACM Press. Reading, MA: Addison-Wesley, 1994.
Jeffries, Ron, Ann Anderson, and Chet Hendrickson. Extreme Programming Installed.
Boston: Addison-Wesley, 2001.
Kemper, Alfons, and Guido Moerkotte. Object-Oriented Database Management. Englewood
Cliffs, NJ: Prentice Hall, 1994.
*Kiczales, Gregor, Jim des Revieres, and Daniel G. Bobrow. The Art of the Metaobject
Protocol. Cambridge, MA: MIT Press, 1993.
Kilov, Haim, and James Ross. Information Modeling: An Object-Oriented Approach. Engle-
wood Cliffs, NJ: PTR Prentice Hall, 1994.
Kim, Won, and Frederick H. Lochovsky, eds. Object-Oriented Concepts, Databases, and
Applications. ACM Press. New York: Addison-Wesley Publishing Company, 1989.
*Knorr-Cetina, Karin D. The Manufacture of Knowledge: An Essay on the Constructivist and
Contextual Nature of Science. Oxford: Pergammon Press, 1981.
Korienek, Gene, and Tom Wrensch. A Quick Trip to Objectland. Englewood Cliffs, NJ:
Prentice-Hall, 1993.
*Lakoff, George, and Mark Johnson. Metaphors We Live By. Chicago: University of Chicago
Press, 1980.
*Lakoff, George. Women, Fire, and Dangerous Things. Chicago: University of Chicago
Press, 1990.
*Lakoff, George, and Mark Johnson. Philosophy in the Flesh: The Embodied Mind and Its
Challenge to Western Thought. New York: Basic Books, 1999.
 Bibliography 315

LaLonde, Wilf. Discovering Smalltalk. Redwood City, CA: Benjamin Cummings Publishing
Co., Inc., 1994.
Larman, Craig. Agile and Iterative Development: A Manager’s Guide. Boston: Addison-
Wesley, 2004.
Lau, Chamond. Smalltalk, Objects, and Design. Greenwich, CT: Manning Publications Co.,
1996.
Lewis, Simon. The Art and Science of Smalltalk. London: Prentice Hall PTR, 1995.
Loomis, Mary E. S. Object Databases: The Essentials. Reading, MA: Addison-Wesley
Publishing Co., 1995.
Lorenz, Mark. Object-Oriented Software Development: A Practical Guide. Englewood Cliffs,
NJ: Prentice-Hall, 1993.
Lorenz, Mark. Rapid Software Development with Smalltalk. New York: SIGS Books, 1995.
Love, Tom. Object Lessons: Lessons Learned in Object-Oriented Development Projects. New
York: SIGS Books, 1993.
*McBreen, Pete. Software Craftsmanship: The New Imperative. Boston: Addison-Wesley,
2002.
McGibbon, Barry. Managing Your Move to Object Technology: Guidelines and Strategies for
a Smooth Transition. New York: SIGS Books, 1995.
McNally, Clayton L., Jr., and Peter Molchan Jr. Micro Focus COBOL Workbench. Boston:
QED Publishing Group, 1993.
Martin, James. Information Engineering: Book 1—Introduction. Englewood Cliffs, NJ:
Prentice-Hall, 1989.
Martin, James, and James J. Odell. Object-Oriented Analysis and Design. Englewood Cliffs,
NJ: Prentice Hall, 1992.
Martin, Robert, Dirk Riehle, and Frank Buschmann, eds. Pattern Languages of Program
Design 3. Cambridge, MA: Addison-Wesley Publishing Co., 1998.
Martin, Robert C., with contributions by James W. Newkirk and Robert S. Koss. Agile Soft-
ware Development: Principles, Patterns, and Practices. Upper Saddle River, NJ:
Pearson Prentice Hall, 2003.
Masini, Gerald, Amedeo Napoli, Dominique Colnet, Daniel Leonard, and Karl Tombre.
Object Oriented Languages. Academic Press. London: Harcourt Brace Jovanovich,
Publishers, 1991.
*Maturana, Humberto R., and Francisco Varela. Autopoiesis and Cognition: The Realization
of the Living. Dordrecht: Reidel, 1980.
316 Object Thinking

*Maturana, Humberto R., and Francisco Varela. The Tree of Knowledge: The Biological Roots
of Human Understanding. Boston: New Science Library, Shambala, 1987.
Meyer, Bertrand. Object Oriented Software Construction. New York: Prentice Hall, Inc.,
1988.
Meyer, Bertrand. Object Success: A Manager’s Guide to Object-Orientation, Its Impact on the
Corporation and Its Use for Reengineering the Software Process. London: Prentice
Hall, 1995.
Miller, Byron. Object Oriented Design Made Easy. Minneapolis, MN: Impatiens Publications,
1993.
*Minsky, Marvin. Society of Mind. New York: Simon and Schuster, 1987.
Montgomery, Stephen L. Object-Oriented Information Engineering: Analysis, Design, and
Implementation. Boston: Academic Press Professional, 1994.
Morris, Derrick, Gareth Evans, Peter Green, and Colin Theaker. Object-Oriented Computer
Systems Engineering. London: Springer-Verlag, 1996.
Mowbray, Thomas J., and Raphael C. Malveau. CORBA Design Patterns. New York: John
Wiley & Sons, Inc., 1997.
Muller, Robert J. Productive Objects: An Applied Project Management Framework. San
Francisco: Morgan Kaufmann Publishers, Inc., 1998.
*Nardi, Bonnie A., and Vicki L. O’Day. Information Ecologies: Using Technology with Heart.
Cambridge, MA: MIT Press, 1999.
Newkirk, James, and Robert C. Martin. Extreme Programming in Practice. Boston: Addison-
Wesley, 2001.
Oliver, David W., Timothy P. Kelliher, and James G. Keegan Jr. Engineering Complex
Systems with Models and Objects. New York: McGraw-Hill, 1997.
Orfali, Robert, Dan Harkey, and Jeri Edwards. The Essential Distributed Objects Survival
Guide. New York: John Wiley and Sons, Inc., 1996.
Page-Jones, Meilir. The Practical Guide to Structured Systems Design, second edition,
Englewood Cliffs, NJ: Prentice-Hall, Inc., 1988.
Palsberg, Jens, and Michael I. Schwartzback. Object-Oriented Type Systems. Chichester,
England: John Wiley and Sons, 1994.
Papurt, David M. Inside the Object Model: The Sensible Use of C++. New York: SIGS Books,
1995.
 Bibliography 317

Parsaye, Kamran, Mark Chignell, Setraqg Khoshafian, and Harry Wong. Intelligent
Databases: Object-Oriented Deductive Hypermedia Technologies. New York:
John Wiley and Sons, Inc., 1989.
Partridge, Chris. Business Objects: Re-engineering for Re-use. Oxford: Butterworth-Heine-
mann, 1996.
Pawson, Richard, and Robert Matthews. Naked Objects. Hoboken, NJ: John Wiley & Sons
Ltd., 2002.
Poppendieck, Mary and Tom. Lean Software Development: An Agile Toolkit. Boston:
Addison-Wesley, 2003.
Pree, Wolfgang. Design Patterns for Object-Oriented Software Development. Wokingham,
England: Addison-Wesley Publishing Company, 1995.
Rasmus, Daniel W. Rethinking Smart Objects: Building Artificial Intelligence with Objects.
SIGS Books. Cambridge: Cambridge University Press, 1999.
Riel, Arthur J. Object-Oriented Design Heuristics. Reading, MA: Addison-Wesley Publishing
Co., Inc., 1996.
Rumbaugh, James, Michael Blaha, William Premerlaini, Frederick Eddy, and William
Lorensen. Object-Oriented Modeling and Design. Englewood Cliffs, NJ: Prentice
Hall PTR, 1991.
Rumbaugh, Dr. James. OMT Insights. New York: SIGS Books, 1996.
Schach, Stephen R. Classical and Object Oriented Software Engineering. Chicago: Irwin,
1996.
Schwaber, Ken, and Mike Beedle. Agile Software Development with Scrum. Upper Saddle
River, NJ: Prentice-Hall, 2002.
Shlaer, Sally, and Stephen J. Mellor. Object-Oriented Systems Analysis: Modeling the World
in Data. Yourdon Press. Englewood Cliffs, NJ: Prentice Hall, 1988.
Shlaer, Sally, and Stephen J. Mellor. Object Lifecycles: Modeling the World in States. Engle-
wood Cliffs, NJ: Prentice Hall, Inc., 1992.
Selic, Brian, Garth Gullekson, and Paul T. Ward. Real-Time Object-Oriented Modeling. New
York: John Wiley and Sons, 1994.
Sessions, Roger. Object Persistence: Beyond Object-Oriented Databases. Upper Saddle
River, NJ: Prentice Hall PTR, 1996.
Stapleton, Jennifer. DSDM: Dynamic Systems Development Method. Harlow, England:
Addison-Wesley, 1997.
318 Object Thinking

*Suchman, L. Plans and Situated Actions. Cambridge, England: Cambridge University Press,
1987.
Succi, Giancarlo, and Michele Marchesi. Extreme Programming Examined. Boston:
Addison-Wesley, 2001.
Sullo, Gary C. Object Engineering: Designing Large-Scale Object-Oriented Systems. New
York: John Wiley and Sons, Inc., 1994.
Sully, Phil. Modeling the World with Objects. New York: Prentice Hall, 1993.
Szyperski, Clemens. Component Software: Beyond Object-Oriented Programming. Harlow,
England: Addison-Wesley, 1999.
Taylor, David A. Object-Oriented Technology: A Manager’s Guide. Reading, MA: Addison-
Wesley, 1990.
Taylor, David A., PhD. Object-Oriented Information Systems: Planning and Implementa-
tion. New York: John Wiley and Sons, 1992.
*Taylor, David A., PhD. Business Engineering with Object Technology. New York: John
Wiley and Sons, 1995.
Tkach, Daniel, and Richard Puttick. Object Technology in Application Development. Red-
wood City, CA: The Benjamin Cummings Publishing Co., 1994.
Vlissides, John M., James O. Coplien, and Norman L. Kerth. Pattern Languages of Program
Design 2. Cambridge, MA: Addison-Wesley Publishing Co., 1996.
Wake, William C. Extreme Programming Explored. Boston: Addison-Wesley, 2002.
Walden, Kim, and Jean-Marc Nerson. Seamless Object-Oriented Software Architecture:
Analysis and Design of Reliable Systems. New York: Prentice Hall PTR, 1995.
Webster, Bruce F. Pitfalls of Object-Oriented Development. New York: M&T Books, 1995.
White, Iseult. Using the Booch Method: A Rational Approach. Redwood City, CA: The
Benjamin Cummings Publishing Company, 1994.
Wilkie, George. Object-Oriented Software Engineering: the Professional Developer’s Guide.
Wokingham, England: Addison-Wesley, 1993.
Wilkinson, Nancy M. Using CRC Cards: An Informal Approach to Object-Oriented Develop-
ment. New York: SIGS Books, 1995.
Williams, John. What Every Software Manager Must Know to Succeed with Object Technol-
ogy. New York: SIGS Books, 1995.
*Winograd, Terry. Bringing Design to Software. New York: ACM Press, 1996.
Winston, Patrick Henry. On to Smalltalk. Reading, MA: Addison-Wesley, 1998.
 Bibliography 319

Wirfs-Brock, Rebecca, Brian Wilkerson, and Lauren Wiener. Designing Object-Oriented

Software. Englewood Cliffs, NJ: Prentice Hall PTR, 1990.
Wirfs-Brock, Rebecca, and Alan McKean. Object Design: Roles, Responsibilities, and Collab-
orations. Boston: Addison-Wesley, 2003.
Witt, Bernard I., F. Terry Baker, and Everett W. Merritt. Software Architecture and Design:
Principles, Models, and Methods. New York: Van Nostrand Reinhold, 1994.
Yourdon, Edward. Object-Oriented Systems Design: An Integrated Approach. Englewood
Cliffs, NJ: Prentice-Hall PTR, 1994.
Zeigler, Bernard P. Object-Oriented Simulation with Hierarchical Modular Models: Intelli-
gent Systems and Endomorphic Systems. Academic Press, Inc. Boston: Harcourt
Brace Jovanovich, Publishers, 1990.
Index
A philosophically based on The Timeless Way of
abstract classes, 133 Building, 60
abstractions, 203, 204, 299–305 pattern language proposed, 143–144
avoiding premature, 299 Timeless Way of Building, The, 60, 144, 163
frameworks, 299–302 animated data entity model, 122
composable document, 300 ant metaphor
described, 299 blind coordination vs. centralized control,
extending, 299 112–113
object routing and tracking, 301 complex adaptive systems, simple elements in,
resource allocation and scheduling, 301–302 113
object thinking vs. XP, 203 traffic signal example, 112
reapplying, 299 anthropomorphism, 93, 101–108, 215, 221–222
aformalism, 160–163 applying object-as-person metaphor, 102–103
importance of human developer, 160–161 delegation, 103
methods as exercises, 163–164 guiding decomposition, 104
determining efficacy of methods, 164 limitations of objects, 103
karma yoga, 163 object-as-agent metaphor, 107
agents, refactoring, 107 objects as COBOL programs, 3
agile development, 18–23. See also extreme objects as actors, 107
programming (XP) Philippe Kahn using musicians to illustrate, 101
practices linked to object thinking, 22–23 replacing metaphors with suppositions,
coding standards, 23 107–108
metaphors, 22 applications
on-site customers, 23 architectures, 263–273. See also architectures
refactoring, 22–23 blackboard, 266
simple design, 22 client/server, 266
predated by object thinking, 31 hierarchical control, 263
values, 19–22 model-view-controller, 264–273
communication, 19 pipes-and-filters, 265
courage, 21–22 artifacts, 248
feedback, 20 assembly specialists, replacing programmers
simplicity, 20–21 with, 307–308
Alexander, Christopher defined, 149, 250
designs as defined by, 249 defining objects in terms of, 251
finding divisions in problem space, 17 designing based on GUIs, 240
gates, passing through to attain mastery, 30 designing GUIs based on, 297–298
mysticism of, rejected by formalists, 61 frameworks, 143
Notes on the Synthesis of Form, 16, 18, 249 minimal-intervention principle, 250
Pattern Language, A: Towns, Buildings, objects
Construction, 60 vs. DBMS execution environments, 149
inspiration for software patterns movement, designing based on GUIs, 297
18 models assuming role of, 269

321
322 architectures

architectures, 263–273 CRC cards, 29

blackboard, 266 criterion for differentiation, 77–78
bulletin board architecture, 267 vs. data, 124
whiteboard architecture, 267 as guide to design, 221
client/server, 266 binding, dynamic (late), 146–147
evocative (object-based), 302–305 disadvantages, 146
adding dynamism to model, 304 support for, in programming languages, 146
essential elements, 304 black box module, equating objects with, 60
purpose, 303 blackboard architecture, 266
frameworks, 143 bulletin board as form of, 267
hierarchical control, 263 whiteboard as form of, 267
model-view-controller, 264–273 Booch, Grady, 2, 10
controllers (coordinators), 270–271 canonical form of systems, 76
models, 269 contribution to UML, 27
outside world, 272–273 breakdown, Martin Heideggers’s notion of,
views, 269 explored by Winograd and Flores, 59
pipes-and-filters, 265 Brooks, Fred
artifacts difficulties in software development, 85–87
defined, 248, 250 object-oriented programming as silver bullet, 85
minimal-intervention principle, 250 business
artificial intelligence (AI) forms, objects required to create, 4
foundation behind research, 50 objects. See components
theory of, 54 process reengineering, 149
artificial life, as challenge to formalism, 53 requirements (stories), 149
Auer, Ken, 241
autopoietec (self-organizing) systems, 59
C
C++
B challenge to Smalltalk, 29
Beck, Kent developed by Bjarne Stroustrup, 38, 41–42
abstractions, when to provide, 203 example of compile-link-test environment, 101
agile development values, 19 focus on computer, 42
coding standards, 23 superseded by Visual Basic and Java, 29
coinventor of CRC cards, 173 charts, state, 181–182
courage advocated by, 21 classes, 130–136
exploration of how metaphor affects XP, 94 abstract vs. concrete, 133
importance of metaphor, 22 combining with is-a-kind-of relationships, 251
maturity levels of XP, 163 as exemplar objects, 131
relationship to Ward Cunningham, 27 hierarchies, 179. See also taxonomies
role in object revolution, 27 libraries, 98, 132–136
Behavior! tool, 274 as object factories, 132
behaviors, 192–193 primitives, using to maintain knowledge,
constraints, 281–290 289–290
implementing, 282, 286 as sets, 131
implementing scripts with methods, 287–288 as storage locations, 131
knowledge maintenance (data) objects, variables, 145
288–290 classification, 203
self-evaluating rules, 282–286 canonical form of systems, 76
 Cunningham, Ward 323

defined, 76 contracts, 234–235

differentiation, criterion for, 77–78 control, centralized
Linnaean taxonomy, 76 failure of Soviet Union, 112
client/server architecture, 266 replacing with distributed cooperation and
Coad, Peter (developer of behavioral approach to communication, 81–83
object development), 76 controllers (coordinators)
COBOL programs event dispatchers, 271
encapsulation in model-view-controller architectures, 270–271
algorithms, 15 traditional thinking indicated by, 11
data structures, 15 convergent engineering, 70
vs. object orientation, 9 coupling, 295
as objects, 4 courage
coding standards, linked to object thinking, 23 confronting fear, 21
collaborates-with relationships, 254 as value in agile development, 21–22
collaboration, 130, 204, 225–226 Cox, Brad
as covert exchange, 130 early advocate of object-oriented programming,
defined, 130 92
encapsulation and, 254 originator of superdistribution concept, 92
graphs, 254 software integrated circuit metaphor coined by,
collagists (application assembly specialists), 92
replacing programmers with, 307–308 CRC cards, 70, 158, 159
collective memory maps (CMMs), 181, 259–263 behavior-centricity, 29
domain-level vs. enterprise-level, 262 compared with object cubes, 201
modeling relationships among data depiction developed by Ward Cunningham, 27, 28
objects, 259–263 elaborated by Rebecca Wirfs-Brock, 173
rules for construction, 262 foundation for object cubes, 28
communication, as value in agile development, invented by Kent Beck and Ward Cunningham,
19 173
components, 142, 218 near disappearance of, 29
composability, 78–81 object cubes derived from, 173
documents, 300 CRUD (create, read, update, destroy) matrices,
Lego bricks, 96 288
requirements for, 78–79 cultures
reuse, 79–81 Greek vs. Roman, 64
byproduct of composability, 150 object, 65–89
refactoring, 80 characteristics, 65
composition cooperating with traditional, 87–89
canonical form of systems, 76 groups included in, 65
defined, 76 value of awareness of object culture, 65
computer science, formalist basis of, 51 Western, shaped by formalist philosophy, 51
computer-aided software engineering (CASE) Cunningham, Ward
tools, 152, 158 coinventor of CRC cards, 173
computers, objects as virtual, 106 CRC cards developed by, 27, 28
computers, thinking like, 184 inspiration for most extreme programming
focus on solution space, 16 practices, 27
isomorphism, 18 relationship to Kent Beck, 27
vs. object thinking, 12–18 role in object revolution, 27
constructivism. See hermeneutics
324 customers

customers designs
models, 5 decision hiding, 40
on-site, linked to object thinking, 23 as defined by Christopher Alexander, 249
GUI-driven
dangers of, 241
D
data vs. object-oriented programming, 110
vs. behavior, 124 of objects, 219–245
as decomposition tool, 76 decisions regarding object design, 221
nonexistent in object thinking, 121 guiding object design by domains and object
as objects, 67, 259–263. See also knowledge, behaviors, 221
maintenance objects implications of thinking toward object design,
poor choice for decomposition tool, 74–75 220
treating as passive, 9 simple, linked to object thinking 22
uniting with procedures, 9 determinism. See formalism
databases development, software, limited concept of reuse,
in ideal object environment, 295 97
impedance mismatch problem, 294–297 diagrams. See also models
management systems, 148 interaction, 175–178
philosophy vs. object thinking, 294 Behavior! tool, 274
reasons for using, 296–297 use cases, 176
relational usefulness of, 178
implications of employing, 45–47 soccer ball
problems with mixing objects and, 294–295 vs. object thinking, 3
datacentricity, 15 reinforcing program thinking, 4
DataItem class, 289–290 static relation, 178–181
DBMS. See databases, management systems class hierarchy diagrams, 179
decomposition, 164 collective memory maps, 181
applying object thinking to, 219 gestalt maps, 179
vs. assembling applications, 98–100 Dijkstra, Edsger Wybe
criteria for proper, 72–75 accomplishments of, 73–74
described by Plato, 72 example of formalist position, 73–74
guided by object-as-person metaphor, 104 discovery, 183–218, 226
object, 8 applying object thinking to decomposition, 219
object thinking heuristics, 283–284 decomposition, 220
Parnas, David Lorge domains
decomposing according to design decisions, anthropology, 186–200
74 understanding, 185–200
On Decomposition (paper written by), 40 object definition, 200–218
providing insufficient information for object abstractions, 203, 204
simulations, 220 capturing object information, 200–218
tools, 74–75 classification, 203
traditional, 8 essentialism, 203
delegation, 213–214 generalizations, 203
vs. multiple inheritance, 139–140 heuristics, 212–218
result of refactoring, 214 vs. object specification, 200
dispatching events, 277–280
 formalism 325

documents, composable, 300 not preserved by databases, 295

domains, 147–149 selecting encapsulating object, 226
anthropology, 186–200 engineering, convergent, 70
behaviors in domain, 192–193 entity models, 5
constructing semantic nets, 189–192 normalization rules, 5
eliminating preconceptions, 192–193 vs. object models, 6
initial focus, 192 vs. UML models, 5
management and, 189 environments
organizing work, 193–200 integrated development (IDEs), 240
defined, 147 visual development (Visual Basic), 240
execution, 148 essentialism, 203
experts, 19 events, dispatching, 277–280
focus of, 78 event dispatchers in model-view-controller
guiding designs by, 221 architectures, 271
implementation, 148 interaction diagrams, 277–278
problem object cubes, 277–278
deriving patterns from, 144 state modeling, 278–280
everything-is-an-object principle applied to, execution domains
70 database management systems (DBMSs), 148
simulating, 71–78 defined, 148
vs. solution spaces, 97, 123 extreme programming (XP), 275. See also agile
as systems, 249–250 development
understanding, 185–200 asserting importance of people, 57
focus of decomposition, 78 culture of, 62
negotiation of domain language, 77 evocative (object-based) architectures, 302–305
Dynabook, idea developed by Alan Kay, 37 incompatibility with Roman thinking, 64
dynamic (late) binding, 146–147 maturity levels, 163, 291
disadvantages, overcoming, 146 object thinking central to, 28
support for, in programming languages, 146 Objectionary, 306–308
dynamic relationships, 273–280. See also static practices, inspired by Ward Cunningham, 27
relationships replacing programmers with collagists
events, dispatching, 277–280 (application assembly specialists), 307–308
scripts resemblance to playing with Lego bricks, 101
interaction diagrams, 274–277 role of metaphor in, 94
XP, 275 systematizing programming practice, 153
dynamism, adding to object-based evocative Universal Vending Machine (UVM), 32
architectural model, 304
F
E feedback, as value in agile development, 20–21
emergence, inconsistency with formalism, 54 Flores, Fernando, 59
encapsulation, 141–142 foreshadowing, 214–215
algorithms, 15 formalism, 50–51, 161–162
COBOL programs, 15 areas of success, 89
collaborating objects and, 254 automated tools (CASE tools), 152
collaboration covert exchange within, 130 basis of computer science, 51
data structures, 15 bias toward, in software development, 25
326 forms, business

formalism, continued based on applications, 297–298

central notions, 51 vs. object thinking, 240–241
challenges to, 53 visual development environments, 240
critiqued by Robert L. Glass, 57 environments presumed in definition of views,
defined, 51, 58 269
equating objects with black box module, 60 focus of early object-oriented programming on,
vs. hermeneutics, 54–58 110
vs. heuristics, 58 GUI-driven designs
methods, 152–153, 160 dangers of, 241
rejecting mysticism, 61 vs. object-oriented programming, 110
Western industrial culture shaped by, 51
forms, business 4
H
Fowler, Martin, Analysis Patterns vs. Design Harel, David (originator of state charts), 181
Patterns (Gang of Four), 144 Heidegger, Martin (hermeneutic philosopher), 59
frameworks, 142–143, 299–302 hermeneutics, 51–53
abstract (foundational), 143 defined, 52
application (vertical market), 143 denial of intrinsic truth, 59
architectural, 143 derivation of term, 52
composable document, 300 vs. formalism, 54–58
described, 299 nondeterminism fundamental to, 53
extending, 299 semantic meaning of documents, 52
implementation, 142 heuristics, 58, 164, 212–218
object routing and tracking, 301 anthropomorphism, 215
objects as core of, 301–302 assuming responsibility for tasks, 213
resource allocation and scheduling, 301–302 avoiding characteristic-specific responsibilities,
function, as choice for decomposition tool, 74–75 216–217
creating proxies for objects, 218
G defined, 58
Gang of Four (GoF), 143–144 delegating responsibilities, 213–214
gestalt maps, 179 determining components, 218
Glass, Robert L. distributing responsibilities, 215
analogy with Greek and Roman cultures, 64 foreshadowing, 214–215
complementing work of Terry Winograd, 59 vs. formalism, 58
critiquing formalism, 57 object discovery, 100
favoring heuristics in software design, 58 object thinking, 283–284
glyphs stating responsibilities, 215
objects, 242–244 hierarchies. See also taxonomies
purpose for using in GUIs, 297 classes, 179. See also classes, libraries
Goldberg, Adele control architecture, 263
author of object behavior analysis (OBA), 154 Hopi language, Whorf-Sapir hypothesis and, 7
influence on Smalltalk, 154 human-derived metaphors
graphical user interfaces (GUIs), 297–298 inheritance, 114–115
building, with Smalltalk IDE tools, 240 responsibility, 115–116
designing humans, as objects, 8, 71
applications based on, 240, 297
 maps 327

I knowledge
idioms, 224–225 maintenance objects, 288–290
determining factors, 224 CRUD matrices, 288
examples of, 224 primitives, 289–290
proper use of, 224 requirements, 221–229
impedance mismatch problem, 149, 294–297 choosing appropriate piece of knowledge,
cause of, 257 225–226
databases, 294–297 responsibilities, 222–224
implementation ways for an object to gain knowledge, 225
domains, 148
languages, 34–36 L
pitfall of moving directly into, 219 Lakoff, George
informalism, 160, 161, 162 research by, revealing central role of metaphor
inheritance, 114–115, 133–136 in human cognition, 93
genealogical charts, 114 Whorf-Sapir hypothesis, 7
multiple, 136, 139–140 languages. See also programming languages
obtaining reuse through, 150 influence on thinking, 7, 8
taxonomies pattern, 143, 163
behavior as criterion for, 114 late binding. See dynamic (late) binding
Linnaean, 114, 133 Lego brick metaphor, 96–101
instance variables, 145 architects as domain experts, 99
integrated development environments (IDEs), compared to software integrated circuits
240 metaphor, 96
interaction diagrams, 175–178 compared to XP, 101
Behavior! tool, 274 delivering objects to end users, 98
specifying scripts with, 274–277 delivering objects to professional assemblers,
use cases, 176 99
usefulness of, 178 employing bricks based on obvious
interfaces characteristics, 99
protocols, 129 Legoland store, events sponsored by, 99
views, as means for interobject communication, small number of simple objects suggested by,
297 100
interpretationalism. See hermeneutics libraries, class. See classes, libraries
is-a-kind-of relationships, 251–253 Linnaean taxonomy, 76, 133
isomorphism, 18 LISP (nonprocedural language), datacentric
development accommodated in, 15
J–K Lorenz, Mark (metrics for object projects), 10–11
Java, C++ superseded by, 29 Low-Income Mortgage Trust (LIMT), 170–173
Kahn, Philippe, 101–102
Kay, Alan M
describing Smalltalk, 43–45 managers, traditional thinking indicated by, 11
Dynabook idea developed by, 37 maps
objects as virtual computers, 106 collective memory, 181, 259–263
thinking about computers changed by, 37 domain-level vs. enterprise-level, 262
Kidd, Jeff (metrics for object projects), 10–11
328 mechanism

maps, collective memory, continued small number of simple objects suggested by,
modeling relationships among data depiction 100
objects, 259–263 linked to object thinking, 22
rules for construction, 262 role in everyday thinking, 93
gestalt, 179 role in XP and object thinking, 94
mechanism. See formalism software as theater, 108–111
messages, 227–240 re-creating old standards, 108
contracts, 234–235 scripting, complexity of, 109
protocols, 227–232 software integrated circuits, 92
state change notifications, 236–240 thinking shaped by, 93–94
types, 128 methods, 159–164
metaphors, 91–116 aformal
anthropomorphism, 93, 101–108 importance of human developer, 160
applying object-as-person metaphor, 102–103 vs. other methods, 161
delegation, 103 behavioral, 158–159
guiding decomposition, 104 difference from traditional methods, 159
limitations of objects, 103 reasons for failure in market, 158–159
object-as-agent metaphor, 107 cultures embracing, 161–164
objects as actors, 107 aformalism, 161–163
Philippe Kahn using musicians to illustrate, formalism, 161–162
101 informalism, 161–162
replacing metaphors with suppositions, data-driven, 155–157
107–108 efficacy, criteria for, 164
ants as exercises, 163–164
vs. centralized control, 112–113 formal, 160
complex adaptive systems, 113 implementing scripts using, 287–288
traffic signal example, 112 informal, 160
central role of, in human cognition, 93 private, 145
contrasted with specifications, 107 public, 145
human-derived software engineering, 157–158
inheritance, 114–115 syncretism, 164–168
responsibility, 115–116 defined, 165
inheritance, 133–136 requirements imposed by object thinking,
multiple, 136 166–167
taxonomies, 133 vs. traditional functions, 145
Lego bricks, 96–101 Microsoft Visual Basic. See Visual Basic
architects as domain experts, 99 minimal-intervention principle, 250
compared to software integrated circuits, 96 models, 168–182
compared to XP, 101 architectural, 263–273
composability of, 96 blackboard, 266
delivering objects to end users, 98 client/server, 266
delivering objects to professional assemblers, hierarchical control, 263
99 pipes-and-filters, 265
employing bricks based on obvious collective memory maps, 259–263
characteristics, 99 domain-level vs. enterprise-level, 262
Legoland store, events sponsored by, 99 rules for construction, 262
customer (entity, object, UML), 5
 object models 329

diagrams O
interaction, 175–178 object analysis, CRC card approach, 27
static relation, 178–181 object applications vs. DBMS execution
entity, 6 environments, 149
in model-view-controller architectures, 268 object behavior analysis (OBA), 154
object object cubes, 173–175, 201–212
vs. entity models, 6 air traffic control example, 209–212
vs. UML models, 6 airplane example, 222–224
object cubes, 173–175 aspects (sides), 173–175
aspects (sides), 173–175 compared with CRC cards, 201
derived from CRC cards, 173 CRC cards as foundation of, 28, 173
of objects self-evaluating rules, 285
animated data entity, 122 object culture, 65–89
soccer ball, 122 characteristics, 65
semantic nets, 169–173 groups included in, 65
as brainstorming tools, 172 value of awareness of object culture, 65
Low-Income Mortgage Trust example, object definition, 200–218
170–173 abstractions, 203, 204
state, 278–280 capturing object information, 200–218
state charts, 181–182 object cubes, 201–212
static relationship, 256–259 stereotypes, 200, 209
UML, 6 classification, 203
model-view-controller (MVC) architectures, essentialism, 203
264–273 generalizations, 203
controllers (coordinators), 270–271 heuristics, 212–218
event dispatchers, 271 anthropomorphism, 215
models, assuming role of application objects, assuming responsibility for tasks, 213
269 avoiding characteristic-specific
outside world, 272–273 responsibilities, 216–217
views, GUI environments and, 269 creating proxies for objects, 218
modularization, destruction of by object delegating responsibilities, 213–214
coupling, 295 determining components, 218
modules, black box, 60 distributing responsibilities, 215
musicians, used by Philippe Kahn to illustrate foreshadowing, 214–215
anthropomorphism, 101 stating responsibilities, 215
vs. object specification, 200
N object discovery, 100
nondeterminism, 53 object methodology, history of, 153–159
nonprocedural languages, 15 behavioral methods, 158–159
normalization rules data-driven methods, 155–157
entity models, 5 Goldberg, Adele, 154
object models, 6 RUP, 155
UML models, 5 software engineering methods, 157–158
Nygaard, Kristen (developer of SIMULA UML, 155
programming language), 37 object models, 5
vs. entity models, 6
vs. UML models, 6
330 object programming (in Smalltalk)

object programming (in Smalltalk), 28 operators, 284

object projects, metrics for, 10–11 outside world, assumed in model-view-controller
object revolution, 27 architectures, 272–273
object routing and tracking framework, 301
object-as-agent metaphor, 107
P
object-as-person metaphor. See Parnas, David Lorge, 163
anthropomorphism decomposition
object-based evocative architecture, 302–305 design decision hiding, 40
adding dynamism to model, 304 using design decisions for, 74
essential elements, 304 head of Strategic Defense Initiative (“Star
purpose, 303 Wars”), 14
Objectionary, 307–308 object thinking vs. computer thinking, 12–18
Objective-C, 92 Pascal, program mode of thinking reinforced by,
object-oriented programming 67
IDE tools, 240 patterns
Smalltalk, 240 architectural, 143
objects, defined, 121 deriving from problem domain, 144
objects, thinking like design vs. implementation, 144
applying old thinking in new contexts, 67 Gang of Four (GoF), 143–144
areas of success of object thinking, 89 languages, 143, 163
autonomy of objects, 81 Martin Fowler vs. GoF, 144
vs. computer thinking, 12–18 reflecting solution space, 144
culture of object thinking, 62 shortcuts for object thinkers, 144
encapsulation, 84 software, inspired by Christopher Alexander, 18
focus on problem space, 16 as solution to recurring programming problem,
vs. GUI-driven design, 110 143
humans as objects, 71 philosophy of implementation languages, 34–36
internalizing object perspective, 30 pipes-and-filters architecture, 265
isomorphism, 18 Plato
prerequisites, 66–83 describing decomposition, 72
composability of objects, 78–81 finding divisions in problem space, 17
primal status of objects, 66–71 polymorphism, 140–141
replacing centralized control, 81–83 postmodernism, 59. See also hermeneutics
simulating problem domain, 71–78 primitives, 289–290
role of metaphor in object thinking, 94 problem domains
software development, 85–87 deriving patterns from, 144
Universal Vending Machine (UVM), 48–50 everything-is-an-object principle applied to, 70
objects, views of, 240–245 finding divisions in, 17
glyphs, 242–244 simulating, 71–78
GUI design vs. object thinking, 240–241 vs. solution spaces, 16, 97, 123
relational database management systems procedures as objects, 67
(RDBMSs), 242 program thinking, 4
separating from objects themselves, 243 programming languages, 38–48
objects, vs. modules, 121–123 bias toward rationalism, 25
 reuse 331

C++ linked to object thinking, 22–23

developed by Bjarne Stroustrup, 41–42 Universal Vending Machine (UVM), 136–139
focus on computer, 42 registrations, accepting for event notification, 126
differences among, 45 relational databases
implementation languages implications of employing, 45–47
philosophy, 34–36 preserving encapsulation of objects, 295
Smalltalk, 98 problems with mixing objects and, 294–295
Visual Basic, 98 management systems (RDBMSs), 242
Java relationships
defining objects according to traditional dynamic, 273–280
thinking, 15 events, 277–280
as imitation of Smalltalk, 154 scripts, 274–277
Objective-C, 92 as objects, 66
reasons for selecting, 33–36 situational, 251
invalid, 33–34 static, 251–273
valid, 34–36 collaborates-with relationships, 254
representing virtual computers, 15 is-a-kind-of relationships, 251–253
shaped by philosophical context, 46 situational relationships, 256–273
SIMULA, 38–41 requirements
focus on problem description, 40 anthropomorphism, 221–222
objectives, 39 applying object thinking to, 219
Smalltalk, 43–45, 154, 240 knowledge, 221–229
described by Alan Kay, 43–45 choosing appropriate piece of, 225–226
designed by Alan Kay, 37 responsibilities, 222–224
IDE tools, 240 ways for an object to gain, 225
Squeak (reinvention of Smalltalk), 37 resource allocation and scheduling framework,
support for dynamic binding, 146 objects forming core of, 301–302
visual, 244–245 responsibilities, 115–116, 123–127, 222–224
programs assigned to objects, 202–212
COBOL air traffic control example, 209–212
vs. object orientation, 9 collaboration, 204
programs as objects, 4 validation, 205–209
size of, and object thinking, 9 assuming, by objects, 213
projects, object, metrics for, 10–11 attributes, 124
Prolog, 15 avoiding characteristic-specific, 216–217
protocols, 129, 227–232 coordinating objects, 127
data-driven vs. behavioral approach, 124
defined, 123
Q–R delegating, by objects, 213–214
quality, software, principles of, 83–87
rationalism. See formalism distributing among objects, 215
refactoring, 80, 204 vs. functions, 124
agents, 107 performing computational tasks, 125
airplane objects, 22–23 stating, 215
delegation arising from, 214 supplying information, 124
extreme programming compared to playing updating objects, 126
with Lego bricks, 101 reuse. See composability
332 routes, in object routing and tracking framework

routes, in object routing and tracking framework, domain-level vs. enterprise-level, 262
301 rules for construction, 262
rules static relationship models (SRMs), 256–259
creating as first-class objects, 286 Smalltalk, 20, 43–45, 101, 154, 240
normalization challenged by C++, 29
entity models, 5 class libraries, 98
object models, 6 computer efficiency subordinate to other goals,
UML models, 5 44
recursiveness of, 285 demise of, 29
self-evaluating, 282–286 described by Alan Kay, 43–45
behaviors, 285–286 designed by Alan Kay, 37
business rules, 283 developed in parallel with graphical user
defined, 5 interfaces, 110
structure, 283–284 development of, 28
RUP (Rational Unified Process), 155 emphasis on GUI building, 240
soccer ball model, 3–4, 122
vs. object thinking, 3
S
scenarios. See interaction diagrams reinforcing program thinking, 4
scripts, implementing with methods, 287–288 software development
self-evaluating rules, 282–286 addressing difficulties, 85–87
behaviors, 285–286 as art, 24
business rules, 283 manufacturing software developers, 24
defined, 5 mastering software development, 24
object cubes, 285 bias toward rationalism, 25
operators, 284 as cultural activity, 25–30
structure, 283–284 failure to remove humanity from, 24
variables, 284 limited concept of reuse, 97
semantic nets, 169–173 as social activity, 26
as brainstorming tools, 172 trends in, 26
compared with static relationship models using rules, 86–87
(SRMs), 257 software integrated circuits, 92
constructing, 189–192 software patterns, 18
Low-Income Mortgage Trust example, 170–173 software quality, principles of, 83–87
semantics, relationship to hermeneutics, 52 software-as-theater metaphor, 108–111
simplicity, as value in agile development, 20 re-creating old standards, 108
SIMULA, 38–41, 164 scripting, complexity of, 109
focus on problem description, 40 solution space vs. problem domain, 16, 97
objectives, 39 Soviet Union, 112
situational relationships, 251, 256–273 Squeak (reinvention of Smalltalk), 37
architectures, 263–273 standards, coding, 23
blackboard, 266 states
client/server, 266 capturing information about UML, 245
hierarchical control, 263 change notifications, 236–240
model-view-controller, 264–273 charts, 181–182
pipes-and-filters, 265 constraints based on, 245
collective memory maps, 259–263 modeling, 278–280
 Visual Studio, as example of visual programming environment 333

static relation diagrams, 178–181 traditionalism. See formalism

class hierarchy diagrams, 179 traffic signals, as example of ant metaphor, 112
collective memory maps, 181 tuples, 267
gestalt maps, 179 types, 146
static relationship models (SRMs), 256–259 typing
static relationships, 251–273. See also dynamic as constrictive mechanism, 69
relationships orthogonality to object thinking, 69
collaborates-with, 254
is-a-kind-of, 251–253
situational, 256–273
U
Unified Modeling Language (UML), 155
architectures, 263–273 capturing state information, 245
collective memory maps, 259–263 contribution of Grady Booch, 27
static relationship models, 256–259 defining objects according to traditional
stereotypes, 200, 209 thinking, 15
stories (business requirements) domain anthropology, 194
development driven by, 124 dominance attained by, 29
used in XP, 149 UML models vs. entity models, 5
Stroustrup, Bjarne (developer of C++), 38, 41–42 UML models vs. object models, 6
structured analysis vs. object thinking, 249–250 Universal Vending Machine (UVM), 32, 48–50
superdistribution, 92 accumulate money story, 94–95
syncretism, 164–168 addressing software development difficulties
defined, 165 with rules, 86–87
requirements imposed by object thinking, make selection story, 104–106
166–167 refactoring, 136–139
systems, domains as, 249–250

V
T validation, 205–209
taxonomies, 251–253 variables
behavior as criterion for, 114 class, 145
derived from classes and is-a-kind-of instance, 145
relationships, 251 self-evaluating rules, 284
Linnaean, 76, 114, 133 views
of objects, 79 as means for interobject communication, 297
rules for, 253 in model-view-controller architectures, 269
Taylor, David A. of objects, 240–245
convergent engineering, 70 RDBMSs, 242
framework for business objects, 70 virtual computers, 15
testing virtual persons, 16
development driven by, 124 Visual Basic
of methods, 287, 288 class libraries, 98
thinking, as influenced by language, 7, 8 superseding C++, 29
Timeless Way of Building, The, 144, 163 Visual C++, as example of compile-link-test
tools environment, 101
automated (CASE tools), 152 visual programming environments, 101
object-modeling, 15 Visual Studio, as example of visual programming
traditional software development, 108. See also environment, 101
formalism
334 vocabularies, specialized

vocabularies, specialized, 117–150 W

auxiliary concepts, 147–150 Western culture, influence of formalist
applications, 149 philosophy on, 51
business process reengineering, 149 whiteboard architecture, 267
business requirements (stories), 149 Whorf-Sapir hypothesis
domains, 147–149 Hopi language, 7
differentiating between object and traditional influence of language on thinking, 7, 8
software concepts, 117–120 widgets
essential terms, 121–129 purpose for using in GUIs, 297
interfaces, 129 types of, 244–245
messages, 128–129 Winograd, Terry
objects, 121–122 complementing work of Robert L. Glass, 59
responsibilities, 123–127. See also influenced by Martin Heidegger, 59
responsibilities Wirfs-Brock, Rebecca
extension terms, 130–144 collaboration graphs, 254
classes, 130–136 contracts, 234
collaboration, 130 CRC cards, 173, 234
components, 142
delegation, 139–140
encapsulation, 141–142
X–Y
XP (extreme programming). See extreme
frameworks, 142–143
programming (XP)
inheritance, 133–136
Yourdon, Edward, 75–76
patterns, 143–144
structured development popularized by, 76
polymorphism, 140–141
using objects to integrate modules, 76
implementation terms, 145–147
dynamic (late) binding, 146–147
methods, 145
variables, 145
About the Author
Currently Dr. David West is a professor in the
School of Business at New Mexico Highlands
University, where he is developing an object-based
curriculum in software architectures, business
engineering, and management information systems.
He also teaches at the University of New Mexico,
where he is engaged in developing a software
development track for students in graduate and
undergraduate computer science.
Prior to joining the faculty at NMHU, he was
an associate professor in the Graduate Programs
in Software at the University of St. Thomas and a
consultant/trainer to several Fortune 500 companies. He has taught courses in
object-oriented development ranging from three-hour introductory sessions for
managers to multiday technical seminars for professional developers, as well as
semester-long courses at both the graduate and the undergraduate level.
He founded and served as the director of the Object Lab at the University
of St. Thomas. The Object Lab was a cooperative effort with local corporations
dedicated to researching and promoting object technology.
He was a cofounder of the Object Technology User Group (the original,
not the Rational-sponsored group), the first editor of its monthly newsletter,
and the principal organizer and chair of two regional conferences sponsored
by OTUG.
Digitalk’s Methods (the first incarnation of Smalltalk for the personal com-
puter, later named Smalltalk/V) was his first object development environment,
used to construct an “automated cultural informant,” a teaching tool for cultural
anthropologists learning ethnographic fieldwork techniques. His object experi-
ence is complemented by more than 20 years of software development work,
ranging from assembly-language programmer to executive management.
His undergraduate education at Macalester College (oriental philosophy
and East Asian history) was capped with an MS in computer science (artificial
intelligence) and an MA in cultural anthropology followed by a PhD in cogni-
tive anthropology. All the graduate degrees were earned at the University of
Wisconsin at Madison.
 Get a Free
e-mail newsletter, updates,
special offers, links to related books,
and more when you
register online!
R egister your Microsoft Press® title on our Web site and you’ll get
a FREE subscription to our e-mail newsletter, Microsoft Press
Book Connections. You’ll find out about newly released and upcoming
books and learning tools, online events, software downloads, special
offers and coupons for Microsoft Press customers, and information
about major Microsoft® product releases. You can also read useful
additional information about all the titles we publish, such as de-
tailed book descriptions, tables of contents and indexes, sample
chapters, links to related books and book series, author biographies,
and reviews by other customers.

Registration is easy. Just visit this Web page

and fill in your information:
http://www.microsoft.com/mspress/register

Proof of Purchase
Use this page as proof of purchase if participating in a promotion or rebate offer on
this title. Proof of purchase must be used in conjunction with other proof(s) of
payment such as your dated sales receipt—see offer details.

Object Thinking
0-7356-1965-4

CUSTOMER NAME
Microsoft Press, PO Box 97017, Redmond, WA 98073-9830