- From: Bijan Parsia <bparsia@cs.man.ac.uk>
- Date: Wed, 16 Jan 2008 20:26:31 +0000
- To: OWL Working Group WG <public-owl-wg@w3.org>
I've added all the missing links in (I hope) a reasonable and consistent way. For each set of slides that Alan gave me, there is a link in the appropriate section of: http://www.w3.org/2007/OWL/wiki/F2F1_Minutes Please, if you gave a presentation, check your session for the slides (and maybe someone elses!) Once Peter checks, we'll have accepted the F2F minutes ;) While doing this I hit two interesting points: One from Boris: ""Every OWL API wants to provide "what are the classes in this ontology", but what does that mean for OWL?""" However, my understanding of the current structural specification is that this is not answered. Ontologies contain axioms, not classes. Are the classes "in" an ontology the classes in the signature of the ontology? All terms? Declared classes? Classes mentioned in imported ontologies but not in local axioms? And in this section: http://www.w3.org/2007/OWL/wiki/F2F1_Minutes#OWL_DL_and_OWL_Full There are these two lines from Jeremy: Jeremy Carroll: for me the whole point is to get compatibility with RDF Jeremy Carroll: A goal is "least surprise" for users of RDF when using OWL This expresses exactly my goals when introducing such features as the skolem reading of bnodes, punning, etc. This is the *goal*. I'm happy to debate the means and sometimes we have to trade off surprises. I think the *general methodology* for achieve less surprise is to look at existing systems and behaviors and try to figure out what people actually expect. The letter or, I would argue, the spirit of the old specs are secondary. Esp. when some aspects of a spec have been neglected or misunderstood or just idle. That's not to say we shouldn't default, all things being equal, to specwards compatibility. But most users don't read specs. I've spent a *lot* of painful time working on understanding the RDF semantics and the various OWL Semantics (including the OWL Full; there are definitely people who understand it better than me, I think, but I've put in comparable effort as anyone). This knowledge and effort could be invalidated by changes. I count that less important than reducing user surprise or improving the user experience. (Goodness knows I've taken on and encouraged others to take on tough implementation challenges! I do want them to be *worth* that effort in terms of the user experience.) So if I get caught up in a "spec implementation technique" as the issue that's an error on my part. For example, I'm convinced that there are details in the "skolemization" proposal that need work (Boris' parse time mapping suggestion has a lot of merit, IMHO). The issue is the user problem (how to handle certain classes of RDF graphs). Cheers, Bijan.
Received on Wednesday, 16 January 2008 20:24:38 UTC