Re: Datatyping

Patrick Stickler wrote:

 >Well, I think you may be reading a bit too much into some recent
 >comments on the rdf-core list. The W3C has a pretty clear process
 >defined and I'm sure the WG will follow it.

Yes, I've been following all that pain.  Sometimes I feel any one of you
could have designed data typed literals better than this committee
effort.  I really hope that you'all end up with something that is simple
and not so riddled with compromises that people don't turn away from RDF
based upon it controted Rube Golberb complexity.

 >Still, one significant question regarding your proposal: What if
 >one defines the range of the age property to be an integer. E.g.
 >   :age rdfs:range xsd:integer .
 >The triple having the lexical node will then not be valid. If you
 >said instead
 >   :age rdfs:range xsd:string .
 >then the triple having the typed node would not be valid.

Why does the MT *need* to make the triple drawn to the LexicalNode
invalid in prescence of a range constraint ?   All I would want the MT
to do, when the age restraint is added, is to entail a new triple drawn
to the  TypedLiteral node, see new graph [2].   Note there is no
constraint placed on an age proerty that says that there must be only
one object of age.  If the author had chosen to place such a constraint
on age (with some appropriate vocabulary), well then he should certainly
not have put an arc in his graph drawn to a LexicalNode;  and if he did,
well then the model theory of that vocabulary should certaily say it was
not valid.


 >One of the key desiderada that has been at the forefront of the
 >datatyping discussions is the compatability between local and
 >global forms of expression (or explicit and implicit forms
 >of expression) and global range assertions.

I understand.

 >I'm presuming here that you intend lexical nodes to be self-denoting.

Yes certainly.

.... thanks for the dialogue.

Seth  Russell

Received on Wednesday, 25 September 2002 13:12:24 UTC