- From: Bijan Parsia <bparsia@cs.man.ac.uk>
- Date: Tue, 17 Oct 2006 14:37:55 +0100
- To: Kendall Clark <kendall@monkeyfist.com>
- Cc: dawg mailing list <public-rdf-dawg@w3.org>
Regrets; I'm being kidnapped by pterodactyls. No! Wait, it's worse than that. I'm withdrawing from the group, and, at the moment, there is no replacement from Manchester, I'm afraid. Basically, I just can't keep up given some other, unavoidable, responsibilities and my current stamina. I certainly learned a lot, which is a good thing. I think I also helped nails some stuff down that wasn't so very nailed down, which is also a good thing. Good luck to getting it done. Hope to meet up with folks at ISWC =============== Few quick thoughts to kinda wrap stuff up: 1) The current semantic framework can be repaired for both its over and under counts. The first (which Jorge pointed out) I think Pat is on top of, the latter (which Eric's been harping on) can be handled (I believe) by extending the notion of a solution to include a justification, that is, a minimal subset of the graph sufficient to entail the binding. (Obviously, you won't *return* these in the normal case, but, in a sense, that's what's actually going on in graph matching and it scales smoothly to description logics; however, for DLs, for non-counting situations, you will want to be a little less restrictive on what systems are forced to return.) 2) I would just adopt the SCS algebra. It's clear, neat, and well done. Jorge I'm sure would be happy to help out. 3) To that end, I would do what many have suggested in having a more abstract language that is the object of the semantics. I would make it so that there was as little implicit as possible. 4) I think the turtle is confusing because of the weakness of the constraints it imposes. OTOH, if y'all do keep it, then I would hope that it stays as closely aligned in *semantic* flavor to turtle and turtle supersets. Essentially, this is another argument for commutativity. 5) If you do go back to graph matching, please be *very* cautious with what you claim for it. There are many issues about applying it to RDF and RDFS (see contradictions!) as a specification mechanism. It would be nice if the text made the interface between the BGP and the algebra very distinct so we could 1) swap in an arbitrary other *specification* mechanism and 2) thus alternative semantics. At OWLED, after ISWC, Kendall and I will be discussing SPARQL with a fairly serious subset of the OWL community: http://clarkparsia.com/weblog/2006/10/13/owled-2006-is-nigh/ http://owl-workshop.man.ac.uk/programme06.shtml Always good to have more! Cheers, Bijan.
Received on Tuesday, 17 October 2006 13:39:00 UTC