Re: Final CFP: XML Database Symposium (XSym03) _at_ VLDB 2003

Mikito Harakiri wrote:
> "Jan Hidders" <jan.hidders_at_pandora.be> wrote in message
> news:8lv3b.6084$HV.28701_at_phobos.telenet-ops.be...

>> Or what about the excellent paper on normal-forms for XML by
>> Marcelo Arenas and Leonid Libkin that got the best paper award at
>> SIGMOD/PODS?

There is only one: you have to impress the programm committee. :-) A meaningless answer, of course, but the point is that there is not really a "they" and this differs per conference and may even differ a bit the next year.

> There
> are so many "interdiscipline" papers! The process goes usually like this:
> "Look ma, wavelets are really cool. What if we apply them to the database
> optimization?" Got a paper, got award.

For PODS it's not that simple. You need at least some deep mathematical results in the sense that they are non-obvious, and you have to show that these have some practical relevance. The program committee member that I know usually take their job very seriously. It seems a bit obvious to me that a paper that breaks new ground gets higher marks then one that presents more of the same.

> In the Information Dependencies case, the really insightful and much more
> well written paper was by Dalkilic&Roberson ([8]). Much less condenced
> material, better examples, much better notation -- overall more pleasure
> to read. Now, application of information dependency to XML normalization
> may be indeed a revolutionary idea (I honestly don't know), but my
> suspicion is that the motivation factor was quite similar: "Look, XML
> really hot, and Information dependencies are cool. What if we marry them?"

That alone would certainly not have been enough. If I would have to summarize their major achievement then it is that they managed to characterize the classical normal forms in an information-theoretic way and they managed to generalize this for the XML data model, which is important because the notion of "update anomaly" is less clear there, and justified that way a normal they had introduced earlier. This has (1) deepened our insight into the classical normal forms for the nested and flat relational model and (2) opened the way for more research on normal forms for more complex data models such as the XML data model.

• Jan Hidders