Re: possible semantic bugs concerning domain and range

Peter F. Patel-Schneider wrote:

>
> It is not possible  to write [integer union string] in N-triples.
>
> I should have been more explicit, and said that this was an XML-Schema
> derived class.

Aha then this is a different issue. Well obviously the problem is due to the
fact that OWL doesn't ... and most likely won't ... understand the nitty
gritty details of XML Schema datatypes.

The issue is that XML Schema union datatypes match lexical to value mappings
in order that they were declared e.g.

xsd:[integer union string] is intended to represent the datatype in which
integer is declared first and hence the lexical value "10" maps to the
integer 10, whereas in xsd:[string union integer] the lexical value "10"
maps to the string "10". I find these sorts of aspects of XML Schema a bit
quirky bordering on downright weird.

>
> > I assume you mean:
> >
> > foo rdfs:range _:x .
> > _:x unionOf (xsd:integer, xsd:string) .
>
> Well, sort of, but the union would be in XML-Schema, not in OWL.
>
... but one approach being touted for a
> while in the RDF Core was to have "the" range of the property determine
> which XML-Schema literal-to-value map to use.  However, this breaks down
> when super-properties can have different ranges (with non-compatible
> literal-to-value maps).
>

Yes, but I (and others) objected to RDF Core doing this for essentially the
fact that one can't know what "the" range of a property is intended to be,
even if there were at most one property ... the lexical to value mapping
would then be different depending on whether the schema were known or not --
and thus non-monotonic.

In any case there is a certain amount of weirdness associated with XML
Schema -- and as you are undoubtedly aware, this weirdness has been ascribed
to the fact that XML Schema was developed in the absense of a formal model,
but in any case I am not sure that we need to wholeheartedly buy into such
quirks -- which is why I'd favor spending more time ensuring that our
datatypes are aligned with the XML Query formalism -- unless you have a
simple fix to all these sorts of issues.

Jonathan

Received on Wednesday, 25 September 2002 20:23:49 UTC