Re: ebXML Representation of Metadata

[Stefan Decker <stefan@db.stanford.edu>]

At 08:22 AM 5/26/00 +0100, Martin Bryan wrote:
> > At this point it might be interesting to note,
> > that RDF is perfectly capable of transporting
> > UML data, since UML is just a specific
> > RDF vocabulary. Please note that RDF is
> > also capable of encoding instance data based
> > on specific UML vocabulary - eg. RDF represents
> > not just a class person, but also a specific person
> > "Stefan".
>
>Being based on a simple tuple model RDF is capable of describing anything. 
>The question is whether or not it is capable of describing anything 
>efficiently in a manner that allows that data to be a) queried in human 
>understandable terms (i.e. not using RDF terminology)  and b) maintained 
>in distributed databases (i.e. can I extract parts the different parts of 
>an RDF stream so that they can be used to update different databases 
>efficientily).

You are right.
Lets compare the two approaches:
To query an RDF-Model for all defined UML-Classes
one just asks for all instances of UML:Class,
which returns another RDF-Model.
An example query with a suitable API (eg. 
http://WWW-DB.Stanford.EDU/~melnik/rdf/api.html )
would look like:

     RDFModel allClasses = myUMLDoc.find("rdf:type","UML:Class").

How would one query an XMI-Document?
Using a SAX-Parser and looking at the events?
Examining the DOM Model of the Document?
This is obviously the wrong level of abstraction.

So one has to recreate most of the
RDF-Datamodel for XMI to be able to pose queries.

But that means (since RDF has the simplest possible
structured datamodel) the XMI-UML-model will be translatable
to RDF very easy (or to put in another way: then XMI is just
a special RDF syntax - which is fine).

Best,
         Stefan



>Martin Bryan