Re: comments on current (6 Nov) draft of RDF MT document

Upon further analysis of the RDF model theory document I realize that some
of my comments are not true.

From: Peter F. Patel-Schneider <pfps@research.bell-labs.com>
Subject: comments on current (6 Nov) draft of RDF MT document
Date: Wed, 06 Nov 2002 23:33:35 -0500 (EST)

> I took at look at the RDF MT document to see what changes needed to be done
> to the OWL MT and found a few problems.
> 
> 1/ The change to make LV a subset of IR has not found its way into several
> places in the document.  In particular, there definition of a simple
> interpretation mentions IR union LV.
> 
> 2/ IP is not introduced in Section 3.1.  Is IP a component of an RDF
> interpretation? 

I missed the introduction of IP in Section 1.4.  It is, roughly, that in a
simple interpretation only elements of IP have property extensions.

> 3/ The condition on IEXT in Section 3.1 does not follow from RDF M&S.  It
> only follows from RDFSS.  In fact, rdf:Resource does not appear at all in
> RDF M&S.
> 
> 4/ The IEXT condition on rdf:Property in an rdf-interpretation is only
> implied by the ICEXT condition on rdf:Property in rdfs-interpretations.

The condition on rdf:Property in an rdf-interpretation (Section 3.1) boils
down to that the class extension of rdf:Property is a superset of IP.
Therefore not all members of the class extension of rdf:Property in an
rdf-interpretation have property extensions and they certainly do not have
to have non-empty property extensions.  

> 5/ The closure condition rdf1 is not valid in rdf-interpretations.
> Therefore, the RDF entailment lemma is false.

From the definition of IP, which I had missed, and from the conditions on
rdf-interpretations any resource that has a non-empty class extension is in
IP, and thus in the class extension of rdf:Property, so the closure
condition is indeed valid, and there is no reason to believe that the RDF
entailment lemma is false.

> 6/ The RDFS closure rules are missing several classes, ranges, and
> subclasses.

This may have been fixed in a more-recent version.

> 7/ The RDFS closure rules are incomplete, even if the above problems are
>    fixed.  Therefore, the RDFS entailment lemma is false.
> 
>    For example, rdf:type rdfs:domain foo . 
>                 a b c .
> 		  RDFS-entails
> 		  a rdf:type foo .
>    because every resource has rdfs:Resource as a type, as I have pointed
>    out before. 
>    Also, a b c .
>          RDFS-entails
> 	   a rdf:type rdfs:Class .
>    because every resource is a subClassOf rdfs:Resource
>    and, rdfs:subClassOf rdfs:domain foo .
>  	  a b c . 
> 	  RDFS-entails
> 	  a rdf:type foo .
>    because 

[I must have deleted part of my previous message before sending it.]

   because every resource is a subClassOf rdfs:Resource
   
> Peter F. Patel-Schneider
> Bell Labs Research

peter

Received on Friday, 8 November 2002 08:28:39 UTC