Re: Using XMLSchema-instance attributes in RDF/XML Syntax

>>>"Peter F. Patel-Schneider" said:
> 
> Dave Beckett <dave.beckett@bristol.ac.uk> said
> > Since I'm the editor of the latter, I'll respond.
> > 
> > Nobody has ever raised it as an issue, that's why it is not there.
> 
> The issue of treating xsi:type specially in RDF/XML *has* been raised
> several times, including my message 

The RDF Core WG issue list doesn't record such an issue on our issues
list but it is related to datatyping which we are working on.

>  http://lists.w3.org/Archives/Public/www-rdf-comments/2002JulSep/0024.html
> 
> xsi:type has been mentioned in at least 69 messages on the w3c-rdfcore-wg
> mailing list, including 
>  http://lists.w3.org/Archives/Public/w3c-rdf-core/2001Oct/0600.html
> and
>  http://lists.w3.org/Archives/Public/w3c-rdf-core/2002Apr/0476.html

Your message above is solely about xsi:type for datatyping literals,
which the RDF Core WG is still discussing on the above list, which I
am on, so of course I've read the messages.

> > Using the W3C XML Schema language (henceforth WXS) to validate
> > RDF/XML is tricky but possible.  ...
> 
> It is?  Where is the XML Schema schema document for RDF/XML then?

Here is how to find it.  Start at the RDF/XML Syntax WD and look in
the section marked:
  Other Syntax Schemas 
  http://www.w3.org/TR/rdf-syntax-grammar/#section-Other-Schemas

where it points to
  XML Schema for RDF/XML
  http://ilrt.org/discovery/2001/09/rdf-xml-schema/

(or typing "XML Schema for RDF/XML" into google gives it as result #3)

which records the state of play - i.e. incomplete.  RDF/XML was
created before WXS was designed.  If they don't work well, you
know who to call.  But see below.

If anyone wants to volunteer to have another attempt at completing
and fixing the above schema, feel free.  I wondered about taking the
RelaxNG schema and converting it.  There are some multi-XML schema
systems out there such as Sun's MXS, but I don't think they can do
this.


> ...  I'm even
> interested in an XML Schema schema document for RDF/XML that only uses a
> fixed collection of properties.

I think I've mentioned my work on that before.

  W3C XML Schemas for simple DC in RDF/XML
  http://ilrt.org/discovery/2002/01/dcxml-xsd/

which has schemas, instance documents (separate) and links to the W3C
WXS validator checking that they work.


> > I've never heard that there was an
> > insistence to scatter xsi:type attributes in RDF/XML data in order to
> > make WXS work.  That sounds like a WXS problem, not RDF/XML's
> > although I'm surprised you can't separate the schema and the instance
> > data.  I've managed to create such things for Dublin Core in RDF/XML
> > with WXS.  I won't go into the other problems WXS has with RDF/XML here.
> 
> Well, I think that it would be extraordinarily useful to be able to say in
> an RDF/XML document that a particular RDF literal had a particular XML
> Schema datatype.  This has nothing to do with making XML Schema work, but
> has to do instead with providing typing for RDF/XML literals.

That's RDF datatyping and RDF Core is working on that.  Right now, as
you know.  That's a different thing again from Thomas' problem.

> > This seems a rather ugly solution to the problem with your schema.
> > Why should just that namespace be ignored?  
> 
> Perhaps because it is the XML Schema namespace?  RDF/XML treats the XML
> namespace specially already.

No, the XML specification treats attributes starting with "xml"
specially.  RDF/XML as revised now uses the XML infoset and all
such namespace issues are handled by XML parsers.  The above-infoset
mapping never sees such prefixes.


> > What about future updates
> > that are bound to happen with WXS?  
> 
> That would be the reason to treat the entire namespace specially.

I meant for new WXS namespaces that are bound to emerge.

> > A good case could be made for
> > ignoring say XHTML's namespaces.  And so on.
> 
> Well, it might be reasonable to ignore XHTML's namespace, reserving it for
> decoration of the document.

So making the ugliness worse.  I'd expect further calls for skipping
other namespaces for a bunch of other useful things that might be
worth sticking in.

<snip/>

Dave

Received on Thursday, 22 August 2002 09:12:12 UTC