- From: Bijan Parsia <bparsia@cs.man.ac.uk>
- Date: Thu, 29 Jan 2009 11:12:20 +0000
- To: W3C OWL Working Group <public-owl-wg@w3.org>
http://www.w3.org/2007/OWL/wiki/LC_Responses/JC1 I've mostly completed a very preliminary draft of a reply to TQ. It's long. It could be longer. It could be shorter and contain more content. One thing I could do is write up my dissection of the story into the wiki and we can point to that. Cheers, Bijan. ----------------------- Dear Jeremy, Thank you for your comment: <http://lists.w3.org/Archives/Public/public-owl-comments/ 2009Jan/0051.html> on the OWL 2 Web Ontology Language last call drafts. The comment you've sent is quite long and complex which attempts to present quite a deep understanding of TopQuadrant's perspective. The working group appreciates that effort and is equally committed to the consensus process. This response is to what we understand as the broadest action we believe you request of us. Where we discerned smaller, specific technical issues that could be sensibly dealt with separately, we separated them out. You will receive distinct responses for each of those. We believe that the fundamental comment and call for action is the following quote: """We ask that many under-motivated new features should be dropped, including all unmotivated new features.""" We perceive the rest of the text as explication of the general approach you would like the group to take when assessing when a feature is under- or un-motivated along with a set of examples of where TopQuadrant would judge a feature to be under- or un-motivated. We distinguish two sorts of judgement: A feature may be under- or un- motivated with respect to TopQuadrant's perception of its current and likely business needs (and of its customer base), and under- or un- motived with respect to a broad enough community (esp. of W3C members) to be worth standardization, all things considered (including potential asymmetric costs to TopQuadrant or to other parties). We believe that is is the latter that is our responsibility to determine to the best of our abilities, though, obviously the former is critical input to those judgments. Essentially, the consensus process is for the WG to take TopQuadrant's input very seriously and to point TopQuadrant to evidence of other parties' interests. Hopefully, we can reach consensus. We welcome ongoing feedback from TopQuadrant. One point of clarification: While we are happy to take your feedback on the LC drafts via comments on the FPWD of the New Features & Rationales(NF&R) document, we wish to point out that since the NF&R document is not complete, there may be significant distortions in your understanding of the motivations, costs, and benefits of the design of OWL2. The working group is satisfied that it did weight the costs and benefits broadly and often made decisions based on minimizing the costs and maximizing the benefits to organizations like your own, often based on feedback from you, Jeremy, personally (which was much appreciated). Thus, we do not think there is sufficient justification to do a systematic re-review of each feature. In particular, you claim that "The rationale document (and the design) has not taken into account the cost of new features particularly to those who do not need them" (I focus on the design issue. The rationale document will be, in due course, updated.) If we examine your illustrative story, we note that is clear that this story could equally well function without OWL 2. For example, one could replace OWL2 throughout with OWL DL and OWL Full and OWL1 with OWL Lite. For syntax, one could have ontologies published in Turtle, NTriples, Manchester Syntax, etc. Furthermore, one could point to extensions like Protege's extensions for QCRs and user defined datatypes and, for that matter, OWL 1.1 and even current versions of OWL. Thus, we do not believe that the story gives new information or a new perspective. One of the goals of OWL 2 from the beginning was to reduce or eliminate, as much as possible, these costs by producing a standard new version to converge on. We believe the overall advantages and, especially, the new clarity of the specification will make it easier for tool developers to cope with real world ontologies and for new tool developers to enter the market. Furthermore, the working group has continually worked to mitigate the transition costs as much as possible. OWL2 deliberately avoid radical new features (such as non-monotonic features, or an entirely new, stratified metamodeling system, or fuzzy extensions). Even features that are well understood and have strong utility and demand were dropped or weakened in response to the sorts of analyses you ask for, e.g., property punning or required n-ary data predicates. While the working group might have erred in some of this, we do not believe that we can make a more accurate prediction at this time, nor do we believe that we did not successfully analyze matters along the way. Regarding bias, we first point out that the working group has had members strongly representing the point of view you advocate, including yourself. If you believe that your interests and comments were not given, procedurally, due consideration, then we encourage you to raise an issue with W3C management. Secondly, members of the working group who might possibly be seen to have the sort of bias you are concerned about are precisely the people who have striven to solicit negative cost analyses. For example, the panel "An OWL Too Far" was proposed by Peter, Ian, Uli, and myself and included Stefan Decker, a long time opponent of OWL DL and, indeed, OWL. (Stefan has not been participating in the OWL 1.1 to 2 effort, so this was a deliberate attempt to bring in a competing voice that had "given up".) Again, the working group, as a collective, could be wrong. Time will tell. But we do not think there is more that we could have done to avoid the problems in methodology that you site. At this point, we just have a difference of opinion. And not a large one, as far as we can tell. TopQuadrant endorses many of the features. The working group believes that they will come to endorse more. There are many features, like property chains, that have been opposed by some people as unmotivated who are now enthusiastic about them. Furthermore, at the moment we have strong evidence from last call of wide endorsement of the overall design.@@pointers to last call comments @@Something about HCLS?!? Finally, we believe that fundamentally reassessing a large number of features -- and even dropping them -- has considerable costs of their own. In general, there hasn't been strong opposition to the feature set and quite a bit of support. Changing that risks breaking the considerable consensus we already have. Without specific evidence of issues, we do not believe that it is sensible, or cost effective, to risk breaking that consensus. As for your other proposal, that we brand the features "Web-SHROIQ". We are rather confused... severing any connection to OWL seems to introduce potentially even more confusion (there's yet another ontology language?). Given the near total overlap between OWL1 and OWL2 (the overwhelming majority of the language is the same; OWL1 ontologies are OWL2 ontologies) while it would certainly make it easier for people who strictly don't need OWL2 features to ignore it, it would also make it exceedingly difficult for everyone else as well as muddling the message. It is also outside the scope of our charter. Please acknowledge receipt of this email and let us know whether or not you are satisfied with the working group's response to your comment. Regards, Bijan Parsia on behalf of the W3C OWL Working Group
Received on Thursday, 29 January 2009 11:09:05 UTC