Re: effects of changes to RDF datatyping

From: Jim Hendler <hendler@cs.umd.edu>
Subject: Re: effects of changes to RDF datatyping
Date: Fri, 16 May 2003 07:49:05 -0400

> At 5:40 AM +0300 5/16/03, Jeremy Carroll wrote:
> >Peter:
> >>  Under the Editor's draft RDF Semantics
> >>  (http:/www.w3.org/2001/sw/RDFCore/TR/WD-rdf-mt-20030117/)
> >
> >>  Because there are no interpretations of
> >>          xsd:int rdf:type rdfs:Datatype .
> >>  that are datatyped with respect to {<rdf:XMLLiteral,XMLLiteral>}.
> >
> >hmmm ...
> >
> >I think this is a comment on that editors draft, and does seem to indicate a
> >bug; but other than that it does not seem relevant to this group.
> >
> >(I will have another look at S&AS section 3 to see whether the same bug
> >applies :) ).
> >
> >Jeremy
> 
> has someone pointed this out to RDF Core?  Discussing their documents 
> on our lists is not totally useful...

I have quite a number of comments on the RDF Semantics, datatyping in
particular, including a comment on the Datatype monotonicity lemma.  I have
performed detailed reviews of multiple versions of the RDF Semantics
document and each time provided comments to either the RDF Core Working
Group or to the editor of the RDF Semantics document.

My views on the semantics of RDF have not been universally accepted within
the RDF Core Working Group.  In an attempt to provide more support for
changes related to the above example that I feel are necessary, I sent the
following message to www-webont-wg@w3.org

> Date: Thu, 20 Mar 2003 14:13:59 -0500 (EST)
> Message-Id: <20030320.141359.133037751.pfps@research.bell-labs.com>
> To: www-webont-wg@w3.org
> Subject: semantics issue with respect to RDF model theory
> From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
> X-Mailer: Mew version 2.2 on Emacs 21.1 / Mule 5.0 (SAKAKI)
> Mime-Version: 1.0
> Content-Type: Text/Plain; charset=us-ascii
> Content-Transfer-Encoding: 7bit
> 
> 
> Currently the RDF model theory says that in a D-interpretation
> that ICEXT(I(rdfs:Datatype)) is a subset of the set of datatypes of D.
> 
> This means that in an XSD-interpretation
> 	xsd:integer rdf:type rdfs:Datatype .
> does not follow from the empty graph.
> It also means that if you add more datatypes, you get more interpretations,
> which violates the datatype monotonicity lemma.
> 
> I believe that there needs to be a different relationship between 
> ICEXT(I(rdfs:Datatype)) and the set of datatypes of D, namely that 
> ICEXT(I(rdfs:Datatype)) is a superset of the set of datatypes of D.
> 
> This would mean that in an XSD-interpretation (and in an OWL
> interpretation) 
> 	xsd:integer rdf:type rdfs:Datatype .
> *does* follow from the empty graph.
> 
> I suggest that the Web Ontology Working Group
> 1/ include the above as a test case,
> 2/ mention to the RDF Core WG that OWL has a test case that violates the
>    RDF model theory, and
> 3/ suggest that the appropriate fix is as I describe above.  (There would
>    have to a non-trivial amount of change required to the RDF Semantics
>    document to make this fix.)
> 
> Peter F. Patel-Schneider
> Bell Labs Research
> Lucent Technologies

This is now in the OWL test cases.  I don't remember any direction from the
Web Ontology Working Group to send an official communication to the RDF
Core Working Group concerning this.

The semantics of datatyping in RDF have undergone radical change since Last
Call.  I have not received any indication that the version in the editor's
draft of the RDF Semantics document have been approved by the RDF Core
Working Group.  I believe that there are more significant changes being
considered by the RDF Core Working Group to RDF datatyping.

Given all this, and given that the Web Ontology Working Group has not been
asked to provide input on the post-Last Call changes, what do you suggest
should be done?


Peter F. Patel-Schneider
Bell Labs Research
Lucent Technologies

Received on Friday, 16 May 2003 08:56:48 UTC