- From: Dan Brickley <danbri@w3.org>
- Date: Tue, 28 May 2002 20:07:35 -0400 (EDT)
- To: patrick hayes <phayes@ai.uwf.edu>
- cc: <w3c-rdfcore-wg@w3.org>
On Tue, 28 May 2002, patrick hayes wrote: > >On Fri, 24 May 2002, Dave Beckett wrote: > > > >> >>>Jeremy Carroll said: > >> > Second a few questions. > >> > > >> > 1. Are there unattached nodes in an RDF graph? > >> > >> No; since we base our describe the rdf graph as triples > >> as node->arc->node rather than any other form. If this is not the > >> case then N-Triples in insufficient to connect the model theory and > >> syntax. > >> > >> > 2. Can any URI ref be a property name or must there be some associated > >> > namespace? > >> > >> Any URI ref. > > > >No, not _any_ URI ref. > > > >Some URI refs don't name properties. > > Well, then any assertion made using them as property names will be > false; but still, its not meaningless, so I think we should allow it. > There might be some reason for someone to treat a phone number as a > property; its not totally crazy. In general, I don't think we are in > a position to say that something definitely does not name a property. Oh, OK, I buy into that v strongly. In _principle_, any URIref might name a property. In practice, many do not. There are ISO object identifiers, for eg., that the digital library community use to name fields in the Z39.50 bibliographic search and retrieval protocol. They could easily (dunno if they do) have URIs, which'd basically be something like iso-obj:<long-number>. Seems very sensible to map into rdf properties, even if no clear notion of derference of those URIs gets us an RDF/XML schema... Dan
Received on Tuesday, 28 May 2002 20:07:35 UTC