Matthew Gertner wrote:
>
> Well, you could have a convention that the property refers to the enclosing
> element. This would limit the power, but I'm not actually convinced that
> that would be a bad thing. One of the greatest things about web-based
> applications, for example, is that they *don't* have Turing machine power,
> restricting the domain and resulting in a (relatively) consistent usage
> paradigm.
If you have that convention then you've blown away the ability for an
element to be graph-connected to another element through a hyperlink. In
other words you've restricted your "graph" to trees!
>...
> If that isn't convincing, you could also attribute the role element used in
> the annotation (the enclosing element could still be the default):
>
> <element which="movie/director">
> <role referent="movie">dc:Creator</role>
> </element>
>
> There's your graph...
Then you have the option of thinking of your graph as conforming to "the
RDF graph model" or "Matthew's graph model." If you choose the former
then you get access to RDF query engines, RDF validators (which will be
much more sophisticated for working with graphs than XML Schema),
RDF-based web application servers like RedFoot, etc. If you choose the
latter then you have to reinvent all of that stuff.
RDF validators will handle subtyping, validating links and probably will
also have some inferencing capabilities.
Paul Prescod
