RE: [Fwd: xmlns, uri+name pairs or just uris..? Clarification n eeded.]

At 10:43 AM 8/1/00 +0100, McBride, Brian wrote:
>Graham,
>
> > It seems reasonable to me that an application doing schema directed
> > processing might have some kind of a priori knowledge of the
> > schemas being
> > used (by embedded labels, implied by the application, or
> > other means), from
> > which the URIs defined by each such schema can be deduced.
>
>For many applications that is no doubt true.  Do you think such
>a priori knowledge would be domain specific?

I would hope that is not required;  i.e. that domain independent 
information could be sufficient.  E.g. a list of namespaces used.

>   The sort of
>application I had in mind was a general purpose tool such as a
>schema directed RDF editor.  Such an editor could be generic, but
>still use schema information such as domain and range constraints,
>labels and comments to assist the user in creating an RDF model.

Good example.  If you know, a priori, a list of namespace URI's used then 
URIs culd be split apart by a form of prefix-matching.

> > >So either the java api needs to change or there needs to be
> > >a way to figure out the namespace.  I guess I'm uncomfortable
> > >with Dan's suggestion of the parser adding statements to the
> > >model - not its job to modify the model it is given really.
> >
> > OK, it's not the parser's job to *modify* the model.  But,
> > for example,
> > additional labels might be defined to be part of the model
> > defined by some
> > given syntax.  Much as having ID on a statement property is
> > sometimes taken
> > to define inclusion of reification of that statement in the
> > resulting model.
>
>Its true that could be done.  Would that require a change to
>the current m&s spec?  Which change would be better - a minor
>fix to the api or to change the spec?

I'm not sure.  Is there an interoperability issue here?  If not, then it's 
a local implemention matter and the spec should probably be silent.

Pierre-Antoinne suggested creating _annotations_ to the model.  If these 
annotations don't need to be communicated then that can be a local 
matter.  If they do need to be communicated then I think they should be 
incorporated into a model spec.

> > >Further, independent of the API, there are situations when
> > it would be
> > >helpful for an RDF processor to be able to determine the
> > >namespace of a resource from its URI, e.g. when it encounters
> > >a subPropertyOf property in a schema.  The processor may well
> > >wish to determine the schema of the super property, e.g. to
> > >determine domain and range constraints.
> >
> > Indeed.  But "in isolation" was part of my comment.  See above.
> >
>
>Sorry, I'm missing something;  I don't understand the significance
>of "in isolation" in this context.  In isolation from what?

 From any other information.

I don't expect to give you an arbitrary URI and ask you which is the 
namespace and which is the local part.  But if I give you a URI and a set 
of allowable namepsace names, I might reasonably expect you to separate the 
parts.

#g

------------
Graham Klyne
(GK@ACM.ORG)

Received on Tuesday, 1 August 2000 13:14:06 UTC