Re: proposal: Structured Datatypes

From: "Jonathan Borden" <jonathan@openhealth.org>
Subject: Re: proposal: Structured Datatypes
Date: Wed, 6 Nov 2002 22:16:18 -0500

> Peter F. Patel-Schneider wrote:
> >
> > Unfortunately, there are several problems with this approach, aside from
> > the problem that Dan has identified.
> >
> > First, the RDF, and thus the OWL, meaning for XML Schema built-in types is
> > incompatible with this meaning.  Second, XML literals in RDF are just
> > literals, not anything else, so there is very limited utility in the
> > scheme.  For example, if the element "a" had type int, <a>010</a> and
> > <a>0010</a> would be different.
> 
> That is the whole point of this. Given the above datatype, and if the
> property is defined as functional the following entailment would hold:
> 
> ex:foo ex:DTprop "<a>010</a><b>aaa</b>"^^rdfs:XMLLiteral
> ex:bar ex:DTprop "<a>00010</a><b>aaa</b>"^^rdfs:XMLLiteral
> 
> =>
> 
> ex:foo owl:sameIndividialAs ex:bar

"<a>010</a><b>aaa</b>"^^rdfs:XMLLiteral and
"<a>00010</a><b>aaa</b>"^^rdfs:XMLLiteral denote different elements of the
RDF domain of discourse.  Thus ex:foo owl:sameIndividialAs ex:bar would not
follow.

> >...Third, this approach would preclude any
> > attempt to do something better, like having the value above be a piece of
> > semi-structured data, containing an int and a string.
> >
> 
> I don't follow, this approach is exactly intended to allow an XML Literal to
> be interpreted as semi-structured (or structured) data, in this case
> containing an int, and string.

But RDF does not do this.  XML Literals in RDF are strings (plus a bit or two).

> That is the whole point, the range of a datatype property is used to provide
> an interpretation of the XML Literal, as more than just simple XML (or a
> base XML infoset) but rather as a PSVI i.e. type adorned infoset.

But, again, OWL can't do this, because OWL can't override the RDF meaning
of XML Literals.

> Now it would have been easier for WebOnt if RDFCore had simply allowed:
> 
> "<a>123</a><b>aaa</b>"^^http://example.org#xType
> 
> but that idea got shot down by RDF Core.

Yes, this would permit doing the right thing, because the denotation of 
"<a>123</a><b>aaa</b>"^^http://example.org#xType could be determined by
http://example.org#xType.  However, the denotation of
"<a>123</a><b>aaa</b>"^^rdfs:XMLLiteral has be fixed by RDF Core, so OWL
can't substitute its own meaning.

> Jonathan

peter

Received on Wednesday, 6 November 2002 23:01:14 UTC