Dan Brickley wrote:
>
> Whatever, I'm far from
> convinced that UML itself is a technology we can simply shoe-horn into
> the Web as-is. I'm thinking eg of the Web architecture axiom that
> everything of value have a URI name on the Web, so it can be
> referenced, described, annotated. UML in itself
> doesn't give Web identifiers to (say) relationship types and entity
> classes. I believe that if UML were tweaked in this direction it would
> start to look a lot like RDF... In this light, Sergey Melnik's work is
> particularly interesting.
>
> Dan
Dan is very right here. I had a bit of a go at trying to do qDC in UML a few months ago and found that it was quite hard to do
it elegantly. The problem is that though it looks like the elements/properties would map to attributes or associations in UML,
you come unstuck when you try to refine the relationships. In fact, both the values and the property-labels are "1st class
objects" - which can be sub-classed, etc - in DC. Sounds like RDF ... For sure you can use UML classes for both and make the
association names rather bland (and suffer something of a combinatorial explosion of classes), but by then I think you've pretty
much lost any advantage of going to UML in the first place.
A weakness that I think UML and RDF share here is that it is not easy in either formalism to say "each of the members of the
class of objects listed in file/registry Y are sub-classes of already known class X" - i.e. importing a (long) list and
asserting that they fit in a particular superclass (e.g. "roles" as specialisations of DC.Contributor).
--
Best regards Simon Cox
CSIRO Exploration and Mining
http://www.ned.dem.csiro.au/research/visualisation/
T: +61 8 9284 8443 F: +61 8 9389 1906 M: 0403 302 672
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|