Re: LANG: A proposal for the layering problem

Dan,

Can you describe what you don't like about my proposal? Is it the fact
that it doesn't build on RDF Schema? If so, let me try to persuade you
otherwise. As time has gone by with this Semantic Web stuff (I've been
doing it for the last five years), I have begun to realize that RDF
Schema is really the thorn in our side that makes everything difficult.
No offense to the working group that designed it, mind you, they had a
good solution to their problem, its just that this approach does not
make it easy to build logical languages on top of it. The problem of
course is the fact that your only building block is tripples. Sure, it's
not impossible to develop ad-hoc methods of getting around this (such as
unasserted tripples), but why bother? We can just create an XML syntax
that clearly and precisely suits our needs. We can even reuse RDF for
the data portion, because that is what RDF is good at.

Let me try to refute the likely arguments in favor of building on top of
RDF schema:

1) RDF Schema is already a W3C standard - But it never reached
recommendation status, indicating that people had serious reservations
about it. However, even if it did become a standard, OWL could be an
alternative approach for describing RDF ontologies for those who want
more expressive power, while RDF Schema can be used by the database
community. You could even have an OWL ontologies and RDF Schemas that
describe the same vocabulary and automate the (partial) generation of
one from the other.

2) RDF Schema tools will have to be changed - So what. DAML+OIL tools
will have to be changed as well if they want to talk OWL. We are too
early in the Semantic Web process to be over-constrained by backward
compatibility.

3) But people are already using RDF - Not a problem. As I've said, I
want to reuse RDF for the data/instance portion of the language. My
proposal is that OWL is an alternative way to describing schemas. The
existing RDF data could be reused by simply creating OWL versions of the
schemas (I can't imagine that there are many of them out there, and like
I said earlier, this can probably be automated rather easily).

If someone has a good argument why we should let RDF Schema shoehorn us
into a suboptimal solution, I'd like to hear it. If not, then the basic
strategy of my proposal should be acceptable.

Jeff

Dan Connolly wrote:
> 
> On Fri, 2002-03-22 at 14:54, Jeff Heflin wrote:
> [...]
> > In
> > terms of Peter and Dieter's "Layering" document, this proposal is
> > probably best described as a diverging sytax and diverging semantics
> > proposal (although it might be a syntactic and semantic extension of
> > RDF).
> 
> I'm not sure I really like this, but I know I prefer
> it to the same-syntax-different-semantics approaches.
> 
> i.e. in this approach, if you mean something
> different, you say it a different way.
> 
> I would find it completely unacceptable if
> two different W3C recommendations gave
> different meanings to the same document.
> 
> --
> Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Tuesday, 26 March 2002 11:22:48 UTC