Review

XML is important, and one should expect to find significant efforts towards implementing XML by mapping it to relational databases. This paper is definitely not the last word on this topic, but it is a start. In fact, this paper appears to me even to be a bit hurried, and I am sure the authors have more definitive work to follow. Yet, there is virtue to being "first-to-market", and this paper is a must-read for any one thinking about implementing XML on top of a relational database.

In particular, I would pay close attention to the list of difficulties listed by the authors -- few other research groups have the expertise and the credibility to catalog these limitations (and have readers accept that these are true system limitations rather than limitations in the ability of less competent researchers to perform the required mappings). The authors do not quite say this, but to me, this long list of limitations clearly indicates that we will not see effective XML implementations in relational (or even object-relational) databases until commercial vendors make serious enhancements to provide effective XML support.