PLDI

The 2006 ACM Conference on Programming Language Design and Implementation (PLDI 2006) was held June 10-16, 2006 in Ottawa, Canada. PLDI 2006 is sponsored by the ACM Special Interest Group on Programming Languages (SIGPLAN), in cooperation with the ACM Special Interest Group on Software Engineering (SIGSOFT). PLDI is a premier forum for researchers, developers, practitioners, and students to present research on programming language design and implementation.

First and foremost, we would like to thank the authors of submitted papers: without high quality input there is no high quality output. We thank the program committee for their commitment to reading, reviewing the submitted papers, selecting the program and providing detailed reviews. Thanks also to the 283 PLDI subreviewers for their input to the reviewing process. Thanks to Jeff Foster of the University of Maryland for organizing the PLDI tutorials. The EasyChair system was used to manage conference submissions and reviews. Thanks to Andrei Voronkov for providing and supporting EasyChair.

The selection of program committee (PC) members followed the standard SIGPLAN guidelines, to achieve a balance across factors such as research sub-disciplines, seniority, gender, geographic location, academia/industry, etc. Submission of papers by PC members was not allowed. Out of 24 invitations to serve on the PC sent out, 19 were accepted. Only one invitee declined because of the no-PC-submission rule.

This year saw a record 174 unique submissions to PLDI (compared to the last eleven years of PLDI). After a number of papers were withdrawn (for a variety of reasons), the PC was left with 169 papers to review. Each paper received three reviews from the PC, which gave each PC member about 26 papers to review. This was quite a heavy load but the committee performed admirably. In addition, a fourth expert outside review was solicited for nearly every paper. A few papers received five reviews.

PC members declared conflicts of interest in reviewing papers following the ACM guidelines, presented to the PC as follows: <i>"Each member of the Program Committee will be responsible for strictly abiding by the rules on conflicts of interest. You are considered to have a conflict of interest on a paper that has an author or co-author in any of the following categories: (1) yourself, (2) your past and current graduate students, (3) your graduate advisors, (4) members of your research group within the last 5 years, (5) a co-author of a paper submitted for publication within the last 5 years, (6) an employee of your immediate organization (academic department, research lab unit, etc.) within the last 5 years, (7) someone with whom you have had a significant funding or financial relationship within the last 5 years, or (8) a member of your family, or (9) someone whose work, for whatever reason, you cannot evaluate objectively."</i>

Papers were graded on a six point scale: 3 (strong accept), 2 (accept), 1 (weak accept), -1 (weak reject), -2 (reject), -3 (strong reject). No zero score (fence sitting) was permitted. Additionally, reviewers provided a confidence score for each paper. Papers that received no positive evaluation score were administratively rejected before the PC meeting.

The PC meeting took place in Charleston, South Carolina on January 14-15, 2006 (after POPL 2006). We first discussed the top 60 papers (in decreasing order of average score). During the meeting, reviewers could propose other papers for discussion. We ended up discussing about 90-100 papers in total at the meeting. Each of the discussed papers was assigned a "champion" who generally had the top score for that paper. The champion summarized the paper's contribution as well as the pros and cons of the paper. The discussion then was opened up to the other reviewers of the paper and to general questions from the rest of the PC. Paper scores were not returned to the authors with the reviews.

We accepted 36 papers, which is a record number of papers for a PLDI program. This reflected the fact that we had a very large pool of quality submissions to choose from, but it is worth saying a few more words about this change.

As PLDI matures, we find it diversifying. In addition to the traditional compiler optimization papers (which did make up the largest category of submissions), we find papers submitted on topics varying from program verification and defect detection to run-time techniques for memory optimization and new programming languages for concurrency. We believe that PLDI benefits from having a diverse portfolio, which a higher acceptance rate enables.

The second point is that by accepting more papers, we increase our chances of finding a "diamond in the rough". If we were to accept only "flawless" papers then we would end up with a program consisting mainly of incremental results in well-established areas. This doesn't help to move our field forward in a significant manner. To grow, PLDI must take some risks. This means we may accept some "slightly flawed" but promising papers in order to expose hidden jewels and to encourage thinking in new directions. Of course, we seek to achieve a high quality technical program. However, "high quality" does not mean "homogeneous".

Because of pre-existing scheduling constraints, the 36 research talks had to fit in two and one-half days, so we shortened talks to 20 minutes plus 5 minutes for discussion (from the usual 25 minutes plus 5 minutes for discussion). Future organizers of PLDI might want to plan for a three-day technical track.