Re: RDFCore WG: Datatyping documents

On 2002-01-28 15:44, "ext Jonathan Borden" <jborden@mediaone.net> wrote:


>> In order for any datatyping scheme to work with RDF, datatypes must
>> have URI denotation. If XML Schema does not define URIs for datatypes
>> (which I think may be debatable), then XML Schema datatypes cannot
>> be used with RDF.
> 
> If this is the case, then _don't_ do it. A halfbaked attempt to shoehorn RDF
> types into XML Schema types is worse than simply defining RDF types. That
> said, I agree that it would be useful if RDF were able to adopt XML Schema
> datatypes. But RDF needs to understand how XML Schema works (by QNames not
> URIs for a start)

We are not looking to "adopt" XML Schema datatypes, per se, but to
allow literals in RDF graphs to be associated with datatypes so that
the interpretation (which value the literal actually denotes) is
clear.

RDF will not have "native" datatypes. In order to support all datatyping
schemes, a generic mechanism, such as URI denotation, is necessary.

Thus, to use XML Schema datatypes in RDF expressions, it must be
possible to refer to them by their URIs.

If that can't be done, cest la vie. Then folks can't use XML Schema
datatypes with RDF.


>> Note that no qnames exist in the RDF graph. Qnames are a mechanism
>> of the XML serialization. If XML Schema datatype qnames cannot
>> be consistently and correctly maped to URIs, then that means
>> that XML Schema and RDF are, at least on a practical level (though
>> certainly not on an abstract level) incompatable.
> 
> That is one issue, the other is that RDF ought not trample on XML Schemas
> URIs and URI prefixes. This simply creates confusion for people reading and
> using the specifications.
> 
> Suppose I were to introduce a new vocabulary and distribute software that
> used new names qualified by the RDF namespace, or something like the RDF
> namespace, would the RDF WG be happy about that?

I think you have misunderstood the usage of the XML Schema qnames
or URIs in the examples. They are intended to simply refer to the
XML Schema defined datatypes, not denote anything else or change
their semantics.
 
>> 
>> However, one final comment, I think that it is possible to work
>> with XML Schema datatype URIs with the present RDF/XML serialization
>> without recourse to qnames. We do, however, tend to use qnames
>> in examples for discussion as a convenience, but could also use
>> only URIs.
>> 
>> The only question remaining, then, is what are the official URIs
>> for pre-defined XML Schema simple datatypes?
>> 
> 
> These are well defined, and if the URIs were limited to these I would have
> no objection. That means (for RDF):
> 
> 1) don't define new URIs (which includes namespace names!) that start with
> http://www.w3.org/2001/XMLSchema
> 
> --- any such URIs need to be documented in XML Schema for _sanity sake_
>
> 2) don't bind the prefix "xsd" to other than that URI, even though it is
> legal according to XML Namespaces, it just causes confusion.
> 
> -- again, this is a sanity check.

If we only used complete URIs and not qnames to refer to XML Schema
datatypes, would that work for you?

That would be doable, I think. Not always convenient, but doable.

Patrick 

--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Monday, 28 January 2002 09:58:20 UTC