- From: Pat Hayes <phayes@ai.uwf.edu>
- Date: Tue, 23 Oct 2001 17:18:32 -0500
- To: Dan Connolly <connolly@w3.org>
- Cc: w3c-rdfcore-wg@w3.org
>Brian McBride wrote: >[...] >> o Pat seems to reckon the model theory is simpler with a set. >>Can we extend >> the notion of a tidy graph so that it removes duplicate >>statements. Any untidy >> graph has an equivalent tidy graph, and the model theory is >>defined in terms of >> that. > >You seem to be discussing this as if issue #rdfms-graph >were to be resolved as per the Sep model theory WD. > >Please see my 11 Oct suggestion to replace it wholesale, which >was greeted with at least two voices of support: > >http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Oct/0186.html Well, since you bring it up, let me add a voice of nonsupport, for several reasons. 1. The apparent simplicity of the definition is largely only apparent, since we have to give the rules for how to understand such a set of triples as a graph. There are many ways to do it, and the only way to be precise is to say what counts as a graph and to give the mapping between the triples syntax and the graph. ( Even if you want to just trash the graph, the same basic issues come up. Do we count two different occurrences of the same literal in two triples as being 'really' the same (tidyness on literal nodes) or not? Etc. If these issues are not resolved in the syntax description they will just come up again, more awkwardly, in the MT itself; the statements of entailment will start to get complicated because they will have to distinguish free and bound symbols, and so on.) 2. As stated it isn't workable. We cannot have 'blank properties' since there would be no way to know which arcs belonged to which properties. The use of nodeID-style labels in the graph really would require that these nodeIDs become incorporated into the core RDF syntax. We could do that, but it seems like a big step to take, and I hope that Frank would object to it as creeping syntaxism. (This is purely a syntax issue, but I think it is important.) 3. Allowing literals on property arcs seems pointless. (The MT could handle it, but unless some datatyping scheme assigns property extensions, any triple with a literal property would be automatically false in every interpretation.) On the other hand, I would suggest that we could go *some* way to liberalizing the syntax. In particular, I can see no rational reason for forbidding literals from the subject position, and lots of very good reasons to permit it, particularly when we consider literal datatyping. Pat -- --------------------------------------------------------------------- IHMC (850)434 8903 home 40 South Alcaniz St. (850)202 4416 office Pensacola, FL 32501 (850)202 4440 fax phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
Received on Tuesday, 23 October 2001 18:36:50 UTC