Re: RDF model theory is now underspecified

From: Dave Beckett <dave.beckett@bristol.ac.uk>
Subject: Re: RDF model theory is now underspecified
Date: Mon, 18 Aug 2003 20:32:44 +0100

> On Mon, 18 Aug 2003 14:50:40 -0400 (EDT)
> "Peter F. Patel-Schneider" <pfps@research.bell-labs.com> wrote:
> 
> > 
> > From: Dave Beckett <dave.beckett@bristol.ac.uk>
> > Subject: Re: RDF model theory is now underspecified
> > Date: Mon, 18 Aug 2003 19:26:08 +0100

[...]

> > > > However, it appears to me that this note means that the RDF model
> > > > theory is now underspecified, as RDF software, including software
> > > > that computes entailments, may choose to treat 
> > > > 	" 3 "^^<http://www.w3.org/2001/XMLSchema#int> 
> > > > as denoting the integer three instead of some non-literal.  
> > > > 
> > > > I view this change as undesirable.
> > > 
> > > We had several implementor feedback reports on this point and made
> > > this change to reflect the reality of running code for XML schema
> > > datatypes.   Implementors now MAY make that entailment above.
> > 
> > This is then a serious source of incompatability between different RDF
> > implementations.
> 
> No, I have already answered that but I can repeat it:
> 
>    We had several implementor feedback reports on this point and made
>    this change to reflect the reality of running code for XML schema
>    datatypes.
> 
> We fixed the problems found by implementors with XSD detail, addressing
> their concern with the changes to RDF Concepts (not in syntax, once I
> revert the mis-paste).

So there were complaints that RDF implementations did something different
from the RDF specification.  The RDF specifications appear to have been
changed so that both the behaviour that had been specified and the
differing behaviour are now allowed.  To my eyes the net result is that the
RDF specifications are now dangerously underspecified.

> > > The RDF Semantics document is not underspecified here, particularly
> > > since the running code we have had feedback on, implements it.
> > > 
> > > Dave
> > 
> > So, does
> > 
> > <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
> >           xmlns:eg="http://example.org/">
> > 
> >    <rdf:Description rdf:about="http://www.example.org/a">
> >      <eg:prop rdf:datatype="http://www.w3.org/2001/XMLSchema#int"> 3
> >      </eg:prop>
> >    </rdf:Description>
> >  
> >  </rdf:RDF>
> > 
> > xsd-entail
> > 
> > <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
> >           xmlns:eg="http://example.org/">
> > 
> >    <rdf:Description rdf:about="http://www.example.org/a">
> >      <eg:prop
> >      rdf:datatype="http://www.w3.org/2001/XMLSchema#int">3</eg:prop>
> >    </rdf:Description>
> >  
> >  </rdf:RDF>
> > 
> > or not?
> 
> You cannot xsd-entail between rdf/xml documents.  RDF datatype
> xsd-entailment is in the RDF Semantics document and happens between
> RDF graphs.

> So the question is meaningless in this form.  The RDF Semantics
> editor can answer it if re-expressed in terms of triples.

OK, so then change the question to

	Does the RDF graph resulting from processing 
	...
	xsd-entail the RDF graph resulting from processing
	...

if that is necessary to get an answer from the working group.

> > In the last-call specification it did not.  Now it appears that the
> > answer is that ``It depends.''  Thus my claim of underspecification.
> 
> There was no answer there to that ill-formed question either
> in the LC drafts.

This is going too far.

The RDF Core specification defines several relationships.  One is between
RDF/XML documents and RDF graphs.  Other relationships the various
entailment relationships between RDF graphs.  It thus makes sense to ask
whether one RDF/XML document entails another and it makes sense to require
that this relationship is well defined.

If this question doesn't make sense, then why is it the subject of the
positive entailment test in RDF Tests between xmlsch-02/test001.rdf and
xmlsch-02/test003.rdf, as well as several other tests?

> The mapping from the RDF concrete (RDF/XML) to RDF abstract syntax
> (RDF triples) will remain well specified, implementable and
> implemented, as already reported to the working group from
> developers.

Well, I wasn't really complaining about the mapping from RDF/XML to RDF
graphs per se.  Instead I was complaining about the effect of a normative
note added to RDF Syntax, as it appears to me to allow for incompatible
behaviour between different implementations of the RDF specification.

> Dave

Peter F. Patel-Schneider

Received on Monday, 18 August 2003 17:02:30 UTC