No more 'Counting Beyond a Yottabyte', or why the W3C process works
Today, at WWW2012 conference, an interesting paper was presented, criticizing a draft design of the SPARQL 1.1 working group, with the controversial title
In their paper, the authors have proven some undesirable complexity results, leading to undesirable performance when evaluating certain path queries in SPARQL1.1 according to the semantics in the Last Call working draft published by the SPARQL Working group in January 2012.
While we acknowledge that scientific papers may use provocative titles, as a more constructive way of feedback the authors have also used the W3C comments mechanism to engage directly with the working group on the issue. The working group has discussed the issue, and arrived at an alternative design of the semantics avoiding the problematic complexity of the original one. The re-design will be part of an upcoming, revised 2nd Last Call working draft of the SPARQL 1.1 Query Language—a proof that the W3C's process works and that feedback, also from outside the working groups, is considered as important to arrive at robust and widely adopted standards.
Apart from property paths, SPARQL 1.1 will add various other new and useful features (such as aggregation, negation, subqueries, updates, entailment regimes, ...) that were demanded by the community and collected in the working group in an initial feature gathering phase. Feedback to the W3C SPARQL working group on its current drafts is welcome to be sent to public-rdf-dawg-comments@w3.org.
I personally think that the process does not exactly work as stated in the above comment, and I see this history more as an example of "a proof that the W3C’s process is not working properly". We have been sending comments by the regular channel (W3C comments mechanism) since Oct 2010 with not too much success. Just to give the reader some examples:
http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2010Oct/0064.html
http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2010Dec/0007.html
http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2011Jan/0023.html
http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2011Mar/0008.html
http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2011Jun/0002.html
http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2011Jul/0004.html
If you look at the process, the SPARQL 1.1 group never discussed the issue with our group. They just received the comment, started an internal discussion, and finally gave a definitive response. This way of handling comments from outside the group is really disappointing. As an example, it is very different to what happened with SPARQL 1.0 where the public list hosted a lot of discussions, and the commenters were allowed to reply, and somehow participate in the internal discussion process. The way SPARQL 1.0 group worked, although not always comfortable for everyone inside the working group, was heavily defended by previous members. This is simply not happening in this new round of standardization, and my personal felling is that the group is under the pressure of some inside members that seems to be more interested in imposing their point of view and having credit for the design than openly discussing all the options and considering the sometimes undeniable scientific evidence that a design decision can be considered wrong.
After all our attempts without being seriously listened by the working group, our only choice was to go for an "unofficial" channel: a scientific conference. But not only that. Since we really wanted to be heard this time, we needed to use a provocative title, and provide a really carefully written paper. Just after publishing a paper in a top conference, and with a provocative title, the group seriously considered our opinion, and actually started a (private) small discussion with us.
So, if the W3C standardization process will still work as in this case, I would say: please we need more 'Counting Beyond a Yottabyte' papers :-)