Re: type/dtype/subclassing and range/subproperties

On Thursday, February 7, 2002, at 05:42 , ext Graham Klyne wrote:

> I finally figured out why 'dtype' is needed, but a corresponding 
> extensions to 'range' is (maybe) not.
>
> 'type' inferences are intimately bound to subclassing, so the addition of 
> subclassing assertions that are quite reasonable with respect to the value 
> domains could completely mess up the datatyping.  (I'm sure you explained 
> this Pat, but somehow I wasn't hearing.)
>
> This did make me wonder is there might not be a similar problem with 
> respect to 'range' and 'subPropertyOf':

Yes. This was discussed awhile back in October
within the context of some
query API binding a literal to a superordinate property
and loosing the datatyping context for correct interpretation.

There was a good bit of discussion between Pat and I on this.
C.f. http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Nov/0176.html

I was myself until just recently unsure of the necessity
of rdf:dtype, but I agree with you that it is needed.

See also my recent query to Pat regarding the relationship
between rdfs:range and rdf:type versus rdf:dtype, which ties
into this issue of the semantics of rdfs:subPropertyOf. We
may in fact need something like rdfs:drange ("deranged"? ;-)
if rdfs:range only implies rdf:type and not rdf:dtype,
otherwise, an application cannot be sure that there is a
valid TDL pairing such that the lexical form is actually
a member of the lexical space of the datatype -- as opposed
to the bNode simply denoting a member of the value space
of the datatype.

It would be good, though, if we could include in the
datatyping summary the "facts" about what rdfs:range,
rdfs:subPropertyOf, etc. refer to (value spaces only).

I try to capture the additional relations between
lexical spaces and canonical lexical spaces in
http://www-nrc.nokia.com/sw/RDFL.html and
http://www-nrc.nokia.com/sw/RDFL.rdf which in some contexts
is useful to know. And an application specific ontology
that uses this datatype definition ontology to define
its lexical datatypes:
http://www-nrc.nokia.com/sw/MARS.rdf

Cheers,

Patrick

Received on Friday, 8 February 2002 04:21:08 UTC