Re: pfps-04 [RDF entailment rules not yet complete]

From: Brian McBride <bwm@hplb.hpl.hp.com>
Subject: Re: pfps-04 [RDF entailment rules not yet complete]
Date: Thu, 31 Jul 2003 16:06:17 +0100

> Peter F. Patel-Schneider wrote:
> > From: Brian McBride <bwm@hplb.hpl.hp.com>
> > Subject: Re: pfps-04 [RDF entailment rules not yet complete]
> > Date: 31 Jul 2003 14:23:01 +0100
> > 
> > 
> >>Peter wrote:
> >>
> >>[[
> >>I believe that the rules for rdf entailments are still incomplete in 
> >>RDF Semantics (Editors [sic] Draft of July 27).
> >>
> >>For example, consider the RDF graph
> >>
> >>	ex:foo ex:bar "<ex/>"^^rdf:XMLLiteral .
> >>
> >>I believe that this graph rdf-entails
> >>
> >>	ex:foo ex:bar "<ex></ex>"^^rdf:XMLLiteral .
> >>]]
> >>
> >>Is "<ex/>" a member of the lexical space of rdf:XMLLiteral?  I don't
> >>think so: 
> >>
> >>http://www.w3.org/2001/sw/RDFCore/TR/WD-rdf-concepts-20030117/#section-XMLLiteral
> >>
> >>[[
> >>The lexical space
> >>        is the set of all strings which: 
> >>              * are well-balanced, self-contained XML data [XML];
> >>              * correspond to exclusive Canonical XML  (with comments,
> >>                with empty InclusiveNamespaces PrefixList) [XML-XC14]
> >>]]
> >>
> >>"<ex/>" does not correspond to exclusive canonical XML; it would have to
> >>be "<ex></ex>".
> >>
> >>Brian
> > 
> > 
> > I am unconvinced.  Just what does ``correspond'' mean here - does it mean
> > ``are'' or does it mean ``resulting in ... when canonicalized''? 
> 
> Neither.  It means that if you encode the string using UTF-8 you get 
> cannonicalized XML.  I'll check this on the WG list.

Huh?   Where does this meaning come from?  I don't see the term anywhere in
Exclusive XML Canonicalization, and the pointer in RDF Concepts points to a
paragraph of Exclusive XML Canonicalization that defines a transformation
from XML document subsets to canonical form, reinforcing the impression
that it is the second meaning that is meant in RDF Concepts.  I don't see
any suitable definition of ``correspond'' in RDF Concepts to override this
indication, as most suitable uses of ``correspond'' simply defer to the
lexical-to-value map of a datatype and the lexical-to-value mapping for
rdf:XMLLiteral ``maps a string to the corresponding exclusive Canonical
XML'', resulting in a recursive loop.

I also note that the lexical space of XML literals in RDF Concepts
(Editors' Draft 28 July 2003) has the condition ``when embedded between an
arbitrary XML start tag and an end tag form a document conforming to XML
Namespaces'' and the value space has the condition that ``when embedded
within an arbitrary XML start tag and an end tag form a document conforming
to XML Namespaces''.  This indicates that the lexical space and the value
space intersect.  Further the productions in XML Namespaces ground out
(following pointers to Appendix B of Extensible Markup Language (XML) 1.0
(Second Edition)) at Unicode characters indicating that the lexical space
and the value space consists of Unicode strings.  This contradicts the Note
that the value space consists of UTF-8 octet sequences.

> Thus "<ex></ex>" is in the lexical space, and "<ex/>" is not.
> 
>   I give it
> > the second meaning, so that "<em></em>" is in the lexical space, but "<" is
> > not.
> 
> I think your conclusions are right, your premise is not.

Please indicate where the RDF documents that contradict my premise.

> > I note that the relevant pointers in RDF Semantics are still broken.  
> > 
> > I also note that if the first meaning is indeed correct, then much of the
> > treatment of XML literals in RDF Semantics needs to be changed.
> 
> Oops - please can you say why.

If the first meaning is correct, then the denotation of a^^rdf:XMLLiteral,
for a in the lexical space of rdf:XMLLiteral, would be a, making much of
the treatment of XML literals in RDF Semantics useless and misleading.

> Brian

Peter F. Patel-Schneider
Bell Labs Research 
Lucent Technologies

PS: It is aggravating that the pointers in RDF Semantics are not correct,
making me over and over hunt for the correct point in RDF Concepts.

Received on Thursday, 31 July 2003 11:36:37 UTC