- From: Sergey Melnik <melnik@db.stanford.edu>
- Date: Sat, 22 Apr 2000 22:49:17 -0700
- To: Dan Brickley <Daniel.Brickley@bristol.ac.uk>
- CC: RDF Interest Group <www-rdf-interest@w3.org>, wchang@adobe.com, connolly@w3.org
Dan, UML Foundation.Core over RDF provides an alternative way of defining a conceptual layer for RDF, not the best possible one though. It is missing/ignoring a number of important features like relationship aggregation, single-tuple vs. multi-tuple cardinality etc. There is a broad spectrum of possibilities of doing things better on top of RDF/RDFS. However, I think we are not quite done with the RDF model itself. For example, I'm not satisfied with the notion of order in RDF. RDF does not specify how ordered relationships are represented. It also does not address n-ary relationships. Still, RDF defines basic language elements like ordinals and typing that allow building those notions on top of it (note that RDFS does not address those issues either). I believe order is so fundamental (compare to the set theory!), that is must go into RDF model (I don't feel as strong about n-ary relationships though. These can be tackled at the conceptual layer). For example, I desperately needed a standard way of expressing order to define the conceptual model of UML on top of RDF. Yes, sure, you can use Bags + ordinals to do the job, but there is a number of possible alternatives for that, so I had to pick one. I think, RDF should specify at least a standard way of expressing ordered relationships. Was this feature also eliminated from the standard somewhere on the way? Please explain. Sergey Dan Brickley wrote: > > This is pretty interesting. As you point out, Walter Chang attempted > something similar, comparing RDF Schema constructs of the time > with similar mechanisms from UML. One thing not immediately apparent > from Walter's note is that RDF Schema has been significantly slimmed > down since then: many of the constraint forms he discussed are not > provided by the RDFS core. > > See the older RDFS spec at http://www.w3.org/TR/1998/WD-rdf-schema-19980409/ > to see what we removed. Basically it was: > > 3.2.1 Necessary and may occur only once : RDFS:ExactlyOne > 3.2.2 Necessary and may occur any number of times : RDFS:OneOrMore > 3.2.3 Optional and may occur at most once : RDFS:ZeroOrOne > 3.2.4 Optional and may occur any number of times : RDFS:ZeroOrMore > > In the working group's judgement, these could be added in later using > the extensibility mechanism. I'm now wondering whether a UML-over-RDF > document might in fact be a mechanism for providing a constraint > utilities vocabulary... > > Dan
Received on Sunday, 23 April 2000 01:38:55 UTC