- From: pat hayes <phayes@ai.uwf.edu>
- Date: Wed, 23 Apr 2003 22:12:33 -0500
- To: Michael Mealling <michael@neonym.net>
- Cc: uri@w3.org
>On Wed, 2003-04-23 at 19:13, pat hayes wrote: >> >>... so how about "a resource is anything that can be referred to". >> > >> >It is synonymous with "a resource is anything that can be identified", >> >> That wasn't my intent. One may be able to refer without identifying, >> which was part of the point. Also, I know what 'refer' means much >> better than I know what 'identify' means. It carries less baggage. > >How can you 'refer' to something if you have no way of talking about >things outside your knowledge other than by a URI? For example, by making assertions, using some URirefs, which can only be true if some things exist, without saying what those things are exactly. Database reasoners and description-logic engines do this all the time, eg something might know that Joe (who has a URI, let us say) has three sons one of whom works for Boeing (which has a URI, say), but may have no URIs for the sons in question. The intelligence community uses software which makes inferences like this, for one example (see the slides at http://www.daml.org/meetings/2001/07/pimeeting.html for example), and so do many industrial applications. There is an entire industry revolving around DL-type inference models and rule inference systems, and these formalisms all have referential semantics as part of their interoperability specification. There are formalisms defined already by the W3C which use URIs to refer, according to semantic rules published as part of the spec of the formalism. >I assume that to >'refer' to something requires some kind of language for doing so? Yes, and the languages which concern me are those defined by W3C specs. >RFC 2396 and URIs have to work and be useful regardless of whether or >not some specific system can 'refer' to things via some other mechanism. Im not talking about mechanisms, but about the semantics of formal languages; and there is no need for the scare quotes, by the way. >I think its useful to point out again that URIs have to work for all >past, present and future systems. That includes things that you would >consider 'on the web', and ones you wouldn't. (My definition of the web >is the set of all RFC-2396-Resources but that's not a common definition >these days). That includes the Semantic Web as well as LDAP, Web >Services as well as VOIP, sip, pop, tip, telnet, etc. I understand, and agree. > >That means that URIs have to work even in the case where two systems >have lethally incompatible models of reality. I agree that there are potential problems here. (There are some slides under 'UWF' at the above website on this very topic, with the title 'when ontologies collide') In fact, any statement to the effect that URIs must identify a unique entity, for example, already produce such lethal incompatibilities. Which is why I was asking for clarification in the first place. >To me that means that >RFC2396bis should say as little as possible. Well, I agree with you there. But 'identify' says a hell of a lot more than 'refer'. >Concepts such as 'refer' >don't belong in there at all..... It just means, designate or identify or indicate or denote or represent.., ie somehow be interpreted to indicate something, in the most general sense possible; that is, with as few extra impositions of meaning as possible. Pat Hayes -- --------------------------------------------------------------------- IHMC (850)434 8903 or (650)494 3973 home 40 South Alcaniz St. (850)202 4416 office Pensacola (850)202 4440 fax FL 32501 (850)291 0667 cell phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes s.pam@ai.uwf.edu for spam
Received on Wednesday, 23 April 2003 23:12:36 UTC