- From: Dan Brickley <danbri@w3.org>
- Date: Mon, 6 Dec 1999 18:38:25 -0500 (EST)
- To: Sergey Melnik <melnik@DB.Stanford.EDU>
- cc: Gabe Beged-Dov <begeddov@jfinity.com>, RDF Interest Group <www-rdf-interest@w3.org>
On Mon, 6 Dec 1999, Sergey Melnik wrote: [...] > That's another tough issue, you are absolutely right. First, on the > model level there no "proper" noname resources, since every resource > must have a URI. Just a brief side-comment for now; longer response another time. Regarding anonymous (ie. 'noname' resources)... I see nothing in RDF model that justifies the claim that we must _know_ the URI string that names each and every resource in a model. We can pass around graphlets that have intermediate anonymous nodes that stand for resources which in other contexts have known URIs. While every resource has to be the 'kind of thing' that might have a URI, ie. an object for which the notion of identity makes sense, my reading of the RDF model is that it remains agnostic on whether implementations must know the URI for all resources in a given graph. Knowing a URI is just knowing some more information about a resource. For example, sometimes we'll know that there is a resource that was the creator of some page, and that the creator had such'n'so age, such'n'so gender, surname etc. But we won't have a URI. If we find out (possibly years later) the URI for the creator of that page, we're still talking about the same resource (entity, object, whatever...). Since a URI is a (particularly intimate) piece of information about a resource, I can well imaging software components knowing facts about some resource without hnowing it's URI. I don't believe RDF model forces us to pretend we know the URI, although I can see how the spec might be read that way. Obviously an implementation will need some way of internally assigning an identifier for such resources, or it would get confused. The only way I can think to deal with this is to reflect URIs themselves into the RDF graph as attributes of resources; sometimes we know these, sometimes we don't. This all boils down to the fact that there are some unresolved issues of Web architecture concerning the relationship between resources and URIs. If a resource can be known by multiple URIs, we get one solution for RDF APIs; if a Web resource has exactly one URI, we get another. In either case it's really a URI and Web architecture issue (though that doesn't mean we can't talk about it here...) Dan
Received on Monday, 6 December 1999 18:38:35 UTC