RDF and XML was Re: [xml-dev] What are Web Services for? ...

John Cowan wrote:


> Paul Prescod scripsit:
>
> > Possibly, but bear in mind that there is a distinction between RDF
> > syntax and model. The RDF people probably have to figure out how to
> > allow ANY arbitrary XML syntax to be mapped into the RDF model. Then it
> > won't be a choice between XML features and RDF features.
>
> Almost true already.  Show me any random XML without mixed content
> in it, and I can interpret it as RDF for you.  It may not make much
> *sense* as RDF, but it will be legal, except for corner cases.
>

Right. There seems to be alot of misunderstanding about the relationship
between RDF and XML. XML is about syntax. The RDF model theory defines a
specific semantics for any RDF/XML i.e. a document serialized in the RDF/XML
syntax. The semantics that is defined relates to the 'graph' that represents
the RDF/XML document. This graph has a precise semantics because what is a
node and what is an arc (or link) is unambiguous and the names of the nodes
and arcs (predicates) are unambiguous. The RDF 'graph' doesn't itself define
the semantics of the URI references that are the names of the nodes and arcs
but at least this is a start (i.e. there can be layers of semantics). Note
that RDF does not use the term 'semantics' in some willy nilly goofy pie in
the sky fashion, rather there is a very precise mathematical definition of
its semantics, namely the RDF model theory (which is in WD).

To define the semantics of URI reference terms, one might use RDF Schema or
more generally OWL (Ontology on the Web Language). One might also use XML
Schema or the XQuery formal language!

All of the above define type hierarchies which at the end of the day are
compatible. In particular the XQuery formal semantics language can define
the type of any particular piece of well-formed XML, so it all works out. By
integrating the XQuery/XML Schema type hierarchies with OWL/RDFS class
hierarchies one can _actually_ define the semantics of any piece of XML.***

I've elaborated more on this approach in:
http://www.openhealth.org/WOWG/IssueStructuredDatatypes or
http://lists.w3.org/Archives/Public/www-webont-wg/2002Apr/0096.html

Jonathan

*** assuming we use the same precise definition of 'semantics'