- From: pat hayes <phayes@ai.uwf.edu>
- Date: Fri, 1 Nov 2002 20:25:53 -0600
- To: w3c-rdfcore-wg@w3.org
Er...guys, Something must have been decided over the summer that I missed somehow. As far as I can see, given the decisions we have made about datatyping, there simply is no useful role for rdfs:Datatype. We don't need it, so why have we got it? As I understand the current decision, the *only* place that a datatype URI can be used so as to exhibit, as it were, its datatyping potential , is inside a typed literal. We aren't going to have any special datatyping entailments which involve the use of a datatype URI as a property (checked in my last email: consensus apparently that have agreed not to do that); there are no long-range datatyping entailments, for sure, so that there is nothing that can be inferred about any interpretations of literals from any reference to a datatype class (that is, something can be said to be in the class, of course, but the fact that it is a *datatype* class has no particular semantic significance for anything else.) And as far as I can see, there is no way to infer that anything is in a datatype class, since the only things that we know are in those classes are literals, which can't be subjects. So we really don't need to impose any special datatyping conditions at all on the property or class extensions of datatype URIs in any interpretation, even in RDFS-with-datatype-entailment. Given this, I can't quite see what Im supposed to say about rdfs:Datatype in the model theory. Like, it's the class of datatypes. But that class isn't defined by RDF(S), and its not even known to RDF(S), so why has RDFS got a special syntax for it? I don't even have any way to refer to it in the MT metalanguage. Time was, when we had datatyping banners unfurled all over the place, that we needed RDFS to be able to 'declare' that some URIref was supposed to be a datatype label, so as to trigger all those range-datatype inferences and so on. But now its obvious from the syntax which urirefs are supposed to be datatypes: the ones inside the typed literals. Saying it in an explicit triple doesn't seem to add anything. And if we say it is, and in fact it isn't (or an RDF engine can't find it) then we are (or the engine is) in just the same pickle as if we had used it inside a typed literal and in fact it isn't. No amount of RDF triple-asserting is going to magically create a non-existent datatype. Unless Im missing something, therefore, I propose that we drop rdfs:Datatype. PROPOSE: do not introduce rdfs:Datatype into the rdfs namespace. Then I can put our very simply boiled-down datatyping into the core RDF MT quite happily, since it won't involve the RDFS vocabulary in any way. It will just be one extra line in the semantics of literals. And if we do keep it, then (unless I am warned of a mistake in the above), I am just going to say that it is like rdf:List, rdf:seeAlso, rdf:first, rdf:rest, rdf:nil and rdf:comment in having no semantics at all. Pat -- --------------------------------------------------------------------- IHMC (850)434 8903 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 Friday, 1 November 2002 21:26:38 UTC