- From: Tim Bray <tbray@textuality.com>
- Date: Wed, 22 Jan 2003 15:37:34 -0800
- To: "Bullard, Claude L (Len)" <clbullar@ingr.com>
- Cc: "'Jeff Bone'" <jbone@deepfile.com>, "Simon St.Laurent" <simonstl@simonstl.com>, www-tag@w3.org
Bullard, Claude L (Len) wrote: > The formalism > that a URI uniquely identifies a "resource", and the > definition of "resource" being nebulous, means that > the URI formal definition does not meet the > requirements of RDF. This same issue has been > repeatedly discussed and is never resolved. A > resource is time-varying and the RDF use isn't. > > In short, it appears that URIs *as formally defined* > are not appropriate for RDF *as formally specified*. This may be a fair statement of where the angst is coming from. Let's assume, as a thought experiment, that it is. So, do we fix the definition of URIs or do we fix the specification of RDF? Since the vast majority of software out there actually does implement URI semantics per RFC2396 (Roy is right; look at a browser or robot source code any time), the cost of changing that definition is higher than the cost of changing RDF, whose deployment is at a much earlier point in the historical curve. But at the moment I'm unconvinced that there really is a serious incompatibility at the *definitional* level. I seem to hear some of the SemWeb people, including TimBL, worrying that inconsistencies in the denotation and use of a URI will break the Semantic Web. I think we all agree that such inconsistencies are bad and reduce the utility of the Web as a whole, and there is language in the most recent Webarch draft stating this. Where I may disagree with the Semantic Web community is that I suspect that such inconsistencies are part of the landscape and there is nothing we can do to make them go away by fiat of any standards body. I expect that to be successful, the RDF-based technologies in the Semantic Web are just going to have to be robust and resilient enough to work around this when it happens. So at the moment I'm just not convinced that it's fruitful to start to write extra rules along the lines of "http:-class URIs must be used to identify things of type yyy" - de facto they're being used as identifiers for the weirdest stuff imaginable and in ways that span the whole spectrum of human creativity, and I just don't see the upside in trying to build fences. -Tim
Received on Wednesday, 22 January 2003 18:37:37 UTC