- From: Dan Connolly <connolly@w3.org>
- Date: Wed, 15 Aug 2001 12:27:33 -0500
- To: w3c-rdfcore-wg@w3.org
We had a Semantic Web Advanced development discussion about the issue http://www.w3.org/2000/03/rdf-tracking/#rdfms-identity-anon-resources I argued earlier that if RDF is to be used to express queries, then Prince nodes must be interpreted ala (exists (?x) (P ?x o)) Recall from Pat's ftf presentation[1] that indeed, if RDF is to be used to express queries, the difference between (exists (?x) ...) and a skolem-constant (aka genid) matters. But we didn't agree that RDF 1.0 was designed to express queries. But from Pat's presentation, it became clear that *any* use of RDF that puts a document "on the pointy end of an entailment arrow" motivates the distinction between existentially quantified variables and genids. i.e. if we want to be able to ask Does rdf doc1 ential rdf doc2? Then we need a model theory with existential quanitification. In Semantic Web Advanced development, this is indeed the case. We need the ability not only to say What are the triples in this document? but also Does doc1.rdf entail doc2.rdf? witness EricM's recent workflow demo http://www.w3.org/2001/07/25-swws/readme.html which relies on TimBL's implementation of the relationship Pat called "instance of"; in TimBL's code (and hence in Eric's stuff) it's called log:includes. There's a pile of test cases in http://www.w3.org/2000/10/swap/test/includes/ [I intended to produce a handful of test cases that don't involve any N3-isms (contexts etc.); just doc1.rdf and doc2.rdf where doc1 is or isn't an instance of doc2. But I spent so much time looking for meeting materials that aren't there that I've got to get onto other stuff just now.] [1] I can't find it online. grumble. -- Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Wednesday, 15 August 2001 13:27:34 UTC