- From: Pat Hayes <phayes@ai.uwf.edu>
- Date: Fri, 16 Nov 2001 13:38:00 -0600
- To: Patrick.Stickler@nokia.com
- Cc: w3c-rdfcore-wg@w3.org
> > >X --s:age--> D --s:inYears--> Y --s:inDecimal--> "12" > >Sergey, or anyone else, > >Can you explain how this representation relates to >data types such as xsd:duration? > >Also, what makes it necessary to have any data typing >properties at all? We can just make eg. s:age a >subPropertyOf xsd:duration and toss in the >literal with no worries. I.e.: > > x s:age "P12Y" . > s:age rdfs:subPropertyOf xsd:duration . > >Eh? > >I.e, I'm unclear on the ramifications of blurring the >distinction between property and data type. > >Is s:age a "general" property or a data type? Both? > >How do we segregate the semantics of 'AGE' and >the semantics of 'DURATION' denoted by s:age? > >It just seems to me that this "have your cake and eat >it" approach is going to bite us in the rear, eventually... Yes, I think Patrick makes an excellent point. All the datatyping proposals on the table use datatype names either as rdf:Class names or rdf:Property names. Both of which raise the question, what makes classes/properties that are datatypes different from other classes/properties? Will an RDF reasoner need to know that a given class/property is a datatype-ish one, instead of an ordinary one? If so, why; and also if so, how is this difference supposed to be expressed in RDFS? Perhaps more acutely: if not, why are we making all this fuss about datatyping? If datatype classes/properties are indistinguishable from normal ones, and can be handled using only normal RDFS reasoning, we don't need to do ANYTHING special about datatypes at all. I would suggest that the authors of the various proposals make their answers clear. Here are mine for P(++); The P(++) proposals require datatypes to be treated specially in the model theory. Datatype classes (and properties) must be distinguishable as such in order for them to be be able to disambiguate occurrences of literals. No other special assumptions, or restrictions on their usage, are required. In the examples given so far this has been done implicitly by the use of particular qname forms such as 'xsd:integer', with the assumption that the required semantic constraints are applied to any URI 'known' to denote a datatype. The P(++) proposals have not included any particular way to make such a distinction in RDFS, but a natural proposal would be to introduce the class names rdfs:datatypeClass and rdfs:datatypeProperty. 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, 16 November 2001 14:39:04 UTC