- From: patrick hayes <phayes@ai.uwf.edu>
- Date: Fri, 31 May 2002 14:43:12 -0500
- To: Dan Brickley <danbri@w3.org>
- Cc: Dan Connolly <connolly@w3.org>, "McBride, Brian" <bwm@hplb.hpl.hp.com>, <www-rdf-comments@w3.org>, <em@w3.org>, <pfps@research.bell-labs.com>, <guha@guha.com>
> > >> > > For example, if RDFS is considered purely as an >> >> RDF vocabulary, then rdfs:subClassOf is not required to be >> >> transitive; >> > >> >??? >> >> Why are you puzzled? There isn't anything in the RDF spec (here >> considered as separate from the RDFS spec) which says that it is >> transitive, and one cannot express transitivity in an RDF graph, so >> how could it possible be required to be transitive? > >'required' is perhaps unhelpful. > >It just *is* a thing that is a transitive relation, just like >phone:+44-117-907wxyz is a thing that is (nearly) my home phone number and >java:org.example.rdf.xyz is a thing that is a Java class and >mailto:danbri@w3.org is my mailbox and isbn:0-88730-824-4 is a book on >marketing and z3950://my.desire.org:2001/ is a Z39.50 bibliographic >database that's currently offline and irc://irc.openprojects.net/rdfig is >my favourite IRC channel. Wrong. RDF (and RDFS and OWL and KIF and every formal assertional language) does not have proper names in it. None of the terms in any of these languages can be said to just *be* any particular thing. The languages only describe; they do not refer. In any case, rdfs:subClassOf isn't a 'thing' in the world in the way that phones and mailboxes are. That term might (indeed, certainly does) refer to different properties in different interpretations. >(Picking another transitive relation that isn't introduced by the RDF >specs might help make this point in a less tangled way. eg: >http://example.com/t1.) > >This is what we get for putting vocabulary constructs (relations, >classes etc) in the domain of discourse. They take on a life of their >own, over and above the descriptions made about them in our various Web >data languages. No, they don't. Ask yourself whether agreeing to talk about unicorns makes unicorns real. >RDF's (possibly contentious) simplifying claim is that some things just >are phones, java classes, mailboxes, books, databases, irc channels, >people, ideas, events, documents, classes, relationship types... Of course that is true: the things themselves 'just are' what they are. The issue is the relationship between those real things and the descriptions of them in a language. In normal language we assume that some of our terms really do refer as proper names: they uniquely identify a particular thing in the real world. But the actual process by which the naming relationship connects names to things is hopelessly complicated, social, indeterminate and generally impossible to specify. Now, you can take the 'scruffy-web' line here (I can almost hear Jim Hendler's voice) and say that yes, the Web really is like that, whoopee. OK, but then don't think that just by saying that life is complicated that we have somehow managed to find the secret of life, or still less that by using a technical term (like "RDF" ) in an informal context we have somehow formalized anything. If that is what "RDF" means then we might as well stop writing specs for it and pretending that it is a well-defined formal language for machine use, because we are just playing silly games. In fact, we might as well just use English sentences as semantic markup: they would convey exactly the same content to both machines and humans. And in any case: you say that this is "RDF's claim' , but where in the spec is that claim made? And suppose that someone were to complain that this claim is wrong, how would we argue against him? >And that >for our purposes we can use similar conventions for making propositions >about all these kinds of things. Just as the limitations of our >descriptive machinery don't affect the actual characteristics of phones, >events, books and databases, they don't affect binary relationship types. >Some are transitive; some aren't. Whether we can express this in some >deep way in machine language is another matter. Most things are unaffected >by their RDF and/or XML descriptions (fortunately!). RDFS can be read as >taking view that RDF classes and properties are unaffected by their >description in RDF Core Schema language 1.0. Properties that are >transitive are still transitive even if RDF Core's Schema language doesn't >provide any way of noting, indicating, flagging this. You are talking about properties as though they were particulars. OK, that is a popular line among mathematicians. Still, what is your basis for claiming that some RDF property name names a *particular* one of these Platonic thingies? If I write <ex:PatsOddProperty> <rdf:type> <ex:PatsClassOfTransitiveProperties> . in RDF, how do you (or anyone) know *which* particular relation <ex:PatsOddProperty> refers to (so that you could check whether it is transitive or not)? I don't think this question is even *meaningful*, myself, let alone has a clear and obvious answer. (In fact, I met rdfs:subClassOf just last week while swimming in the Gulf of Mexico, and it didn't look at all transitive at the time.) Pat -- --------------------------------------------------------------------- IHMC (850)434 8903 home 40 South Alcaniz St. (850)202 4416 office Pensacola, FL 32501 (850)202 4440 fax phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
Received on Friday, 31 May 2002 15:43:27 UTC