Re: proposal to close Issue 5.8 Datatypes

From: Jim Hendler <hendler@cs.umd.edu>
Subject: Re: proposal to close Issue 5.8 Datatypes
Date: Thu, 21 Nov 2002 21:40:37 -0500

> >5/ Cardinality restrictions in the exchange syntax for OWL will use typed
> >    literals, as in
> >              _:x rdf:type owl:Restriction .
> >              _:x owl:onProperty ex:foo .
> >              _:x owl:cardinality "5"^^xsd:decimal .
> 
> Peter-
>   you were strong on the phone today with this, where some people 
> seemed to think going with cardinality as numerals was better for 
> various reasons.  I actually have no idea which is better, but the 
> numerals one would be closer to D+O (i.e. not require recoding all 
> the cardinality statements and rewriting all the parsers).  Can you 
> outline why you felt so strongly about this so the rest of us can 
> understand the argument?

There are several problems in using untyped literals.
1/ The syntax and semantics would have to have to define which Unicode
   strings are allowable as numerals, and the semantics would have to
   include the mapping from such numerals to integers.  DAML+OIL does not
   address these issues.
2/ RDF untyped literals are strange beasts.  How would we handle the
   language tag, for example?

I do not believe that DAML+OIL uses numerals for cardinalities.  The
DAML+OIL reference document claims that cardinalities are non-negative
integers, not numerals corresponding to non-negative integers.  The model
theory document assumes that cardinalities are numbers, not numerals.  The
axiomatization states that the range for cardinalities is
NonNegativeInteger.  The RDF Schema for DAML+OIL states that the range of
cardinalities is xsd:nonNegativeInteger.

This sure looks as if DAML+OIL cardinalities are non-negative integers, not
numerals.  As the (only) way of incoporating non-negative integers into RDF
graphs is via typed literals, to retain compatability with DAML+OIL the
exchange syntax for OWL should use typed literals for cardinalities.  

If RDF allowed long-range typing, then OWL cardinalities could use that but
RDF Core has decided not to include long-range datatyping.  Perhaps WebOnt
should protest that RDF Core should have allowed long-range datatyping.

Or, maybe, WebOnt should protest that RDF Core should allow XML Schema
typing to be used in RDF.  This would allow, I think, the ability to skip
the explicit typing that will otherwise be required for OWL cardinalities.

>   thanks
>   Jim H.

> p.s. Also, in your example why do you propose decimal rather than 
> integer for cardinality (or positive integer)? Is this important? 
> Would we insist is is any typed literal, or only the correct ones for 
> cardinalities?

The RDF meaning of "5"^^xsd:decimal is the same as the meaning of
"5"^^xsd:integer, so either can be used.  (Well, not quite, as I think
there is a bug in the MT for RDF datatyping, but I think that this will be
appropriately fixed.)  Acually, of course, only non-negative integers are
allowed, so the normal datatype would be that one.

peter

Received on Thursday, 21 November 2002 22:08:47 UTC