- From: Dan Brickley <danbri@w3.org>
- Date: Sat, 10 Apr 2004 00:45:30 -0400
- To: Pat Hayes <phayes@ihmc.us>
- Cc: Dan Connolly <connolly@w3.org>, public-sw-meaning@w3c.org
* Pat Hayes <phayes@ihmc.us> [2004-04-09 23:19-0500] > > >I think I get it now... > > > >On Fri, 2004-04-09 at 16:19, Pat Hayes wrote: > >I just scanned > >the document, and nowhere does it say "each URI denotes a unique > >resource"... at least not using the word "unique". I'll have > >to re-read your message to public-webarch-comments, where > >you excerpt lots of text that bothers you. > > > >I'm only aware of 1 at this point: > >We still have "A URI must be assigned to a resource in order for agents > >to be able to refer to the resource" which overstates the case since > >you can refer to things using owl:InverseFunctionalProperty expressions > >but without giving them URIs. > > > > Maybe this could be (perhaps later in the document) modified or > expanded to something like > > "SW formalisms such as OWL provide ways to refer to resources > indirectly by describing their properties. This extends the reach of > the Web, by making it possible to refer to resources which do not > have a URI assigned to them but are suitably related to resources > identified by URIs." Yes, I like this direction. We should be careful also to allow the possibility that the reference-by-description can be done with Literal-valued properties, as well as URI-valued properties (and for that matter, bNode-valued properties, if such scenarios can plausibly be concocted). Literal-valued properties are very useful for this, I think. <c:Company c:nasdaqCode="MSFT" c:name="Microsoft"/> and suchlike can be useful packets of information, without using URIs for anything beyond the terms in the ontology. So I would change ending of your last sentence towards something like: "...are suitably related to resources that are themselves unambiguously identified through appropriate descriptions or URIs" (too wordy to actually use, but in that vein ok?) > > maybe with a disclaimer about this being relatively new and not fully > explored technology yet, or some such. Yes, I believe (in light of the 3-4 year FOAF experiment) that such a disclaimer is appropriate. When we started FOAF there was no class owl:InverseFunctionalProperty. Now we have it, but I am not 100% convinced the semantics it has (due to the approach we took to formalising RDF and OWL) capture entirely what we need for reference-by-description. Specifically, OWL allows that the property/value pair might match different individuals in different interpretations, and guarantees "at most one"-ness only with regard to a single interpretation. In FOAF, I say that properties such as foaf:mbox and foaf:homepage are "Static inverse functional properties", as a (not yet formalised) way of claiming that they cannot take different values at different times. If you believe "a1 foaf:homepage d2", you should not be prepared to believe a2 foaf:homepage d2" at a later date. I don't believe these issues are fully explored yet, so am wary of sending an "OWL solves this once and for all" message. > That would both qualify the > overstatement appropriately, and also acknowledge that 'descriptive' > reference does ultimately depend on some kind of external anchoring > and can't be done *just* by describing things. Yep > > > >> You can't have it both ways. If the Web is essentially a trade in > >> descriptions, then unique URI reference is achievable only by some > >> external technique of rigid designation, many URIs are not unique > >> identifiers, and most 'resources' are not uniquely designated or > >> identified. Which is fine and fits OWL/RDF. > > > >Yes, quite. > > > >> Or, if URIs really are required to be unique designators, ie names, > >> then there needs to be some principled way to baptize resources with > >> URIs. Which would be fine also. And it would be fine to have some > >> URIs one way and others the other way, as long as we could tell which > >> was which for most purposes (which I think is close to the actual > >> situation, in fact). But right now we seem to have a reality which is > >> descriptive, or a mixture, and a rhetoric which is insistently > >> nominal, and they don't fit together. Which is not fine. > > > >Hmm... I guess I think of the architecture as being ideally nominal, > >but practically descriptive. That is a nice way of putting it. BTW the recent RDF-in-XHTML progress raises prospect of reference-by-description at the level of hypertext markup in human readable documents. This would/will/could be fantastic for data merging and aggregation applications. Instead of writing: <p>I went to the <a href="http://w3cfood.example.com/">W3C canteen</a> for a delicious meal</p>, you'd write some expression (I forget the syntax) that makes it clear that the prose fragment has as its primaryTopic a "Restaurant" that has a homepage of [...] and maybe a phoneNumber of [...]. Same for movie reviews etc. With some UI support (eg. in weblogging tools) we could get quite a lot more clarity of meaning in hypertext prose. Note that this still works regardless of any worrying about formal OWL and reference-by-description issues. Our discussions there correspond only to the distinction between saying "a" and "the", which I hope could be clarified through annotating the properties/classes involved, without change to instance data aspects of the problem. Dan
Received on Saturday, 10 April 2004 00:45:30 UTC