- From: Bill dehOra <BdehOra@interx.com>
- Date: Thu, 18 Jan 2001 10:18:14 -0000
- To: "'Alex Kuijper'" <A.Kuijper@InfoRay.NL>, www-rdf-interest@w3.org
> I have a question about multi occurrences of the same > predicate to the same > object > (resource or literal). RDF syntax does not say anything about that. > [...] > How should a RDF parser handle this? Should the multi > occurrences be added > to the graph or skipped? A parser shouldn't handle this, a parser handles syntax (at a glance this is syntactically legal RDF). However it might be semantic nonsense in a domain of discourse. Traditionally this issue of domain specific nonsense belongs at either end of an RDF transport. So if I had a data model or schema which specified that a resource should have one and only one given predicate, then a validating process can be built to trap such an infraction and/or prevent such an infraction being passed to a serialiser in the first case (well, I can dream). So (and deserialisation realistically implies parsing xml): App [SCP] -> rdfserializer -> on the wire -> rdfdeserializer -> [SCP] App2 SCP stands for "semantic checkpoint", which is where you're traditionally looking for semantic nonsense, i.e. the semantics are held in endpoint "codecs" that were built and agreed upon out of band. I think some folk here would suggest that it's possible to logically (which is not to say actually) tunnel further rdf that would semantically validate the above rdf, i.e. keep as much of the semantics as possible in band and make the endpoints less brittle and highly decoupled. So your codecs I guess morph into general purpose evaluators of utterances. I have an imagining that, architecturally, this is part of what it means to have a semantic web, but I've never heard it put as such. Any semantic web visionaries care to comment? -Bill ----- Bill de hÓra : InterX : bdehora@interx.com
Received on Thursday, 18 January 2001 05:19:02 UTC