facets, XSD, SCD, external datatypes (was Re: A brief primer on Qnames and URIs)

Bijan Parsia wrote:
> 
> On 15 Nov 2007, at 10:45, Peter F. Patel-Schneider wrote:
> 
>> Summary:  There are approved URIs for the XSD facets.
> 
> Ok, nice. Thanks for verifying that.
> 
>> Result:  I think that thus we can use the XSD names for the XSD facets
>>      in OWL 1.1.
> 
> Welllllll.

I am convinced.

> 
> [snip]
>> Note, however, that the URIs for the datatypes are *not* the
>> concatenation of their namespace http://www.w3.org/2001/XMLSchema and
>> their local name.  For example the URI for integer is
>> http://www.w3.org/2001/XMLSchema#integer, not
>> http://www.w3.org/2001/XMLSchemainteger.  One would thus expect that in
>> an OWL or RDF document that the namespace for prefix xsd would be
>> http://www.w3.org/2001/XMLSchema# and not the XML Schema namespace.
> [snip]

This problem has been there from the beginning; everyone's learnt to 
live with it; the facets don't change the problem with the namespace #.

> 
> Ye-ouch. So if we used qnames that could turn into the right uris, we 
> wouldn't be reusing their qnames, in fact, we wouldn't be using 
> universal names *in their namespace*, just one very similar that they 
> probably still own.

We don't need to ask, but we could let them know we were doing this.

> 
> My head hurts :(

I suggest a massage :)

> 
> So, at the least, we still need, for XML Syntaxes, to confirm that for 
> xmlns:xsd="http://www.w3.org/2001/XMLSchema#" that things like
> 
>     <xsd:minInclusive....
> 
> or
>     <... xsd:minInclusive="...
> 
> are cool with them. I would hope that it is!
> 
> OTOH, if we adopt their syntax for inline derived types, then we need 
> *two* xsd declarations (the inner one shadowing the outer one)?!?!
> 
> I have to say that I would be willing to support a more user rational 
> scheme over the objections (if any) of the XML Schema working group. To 
> wit:
>     We need to have resolvable URIs for user defined datatypes in XML 
> Schema documents. If they punt, we should take over.
>     We need sensible names for facet elements and URIs. If they 
> disagree, we should take our case upwards.
>     We need reasonable syntax for inline datatypes. I'm open that that 
> would be fragments of XML Schema, but given the trickinesses of XML 
> Literals in RDF/XML, I'm also open to something like the current design.
> 
> I've no evidence at the moment that the current XML Schema WG wouldn't 
> be friendly and helpful on this points, but I strongly disagree with 
> Jeremy's expressed opinion in the telecon that "If the XML Schema WG 
> doesn't solve this, then it remains unsolved". *If* they are 
> recalcitrant, we should still solve our problems.


With Peter's e-mail, I am confident that we can have:

- in-line XSD datatypes with our own syntax
- in-line XSD datatypes using their syntax (but probably not in RDF/XML; 
the xml literals in RDF/XML are defined via exclusive XML 
Canonicalization, that is not compatabilte with shipping XML Schema 
fragments)
- external references to XSD datatypes with an explicit id attribute
   [Hence allowing an ontology to be shipped in two files, one 
containing all the datatypes]

The only missing desiderata was external references to arbitrary XSD 
datatypes, which does depend on XML Schema's SCD doc.
I personally think we should be prepared to cut our losses at that point.

Jeremy

Received on Thursday, 15 November 2007 12:49:42 UTC