Re: UPDATE: document on layering OWL on top of RDF

> > Benefits:
> > 
> > 1/ RDF tools, including parsers, database managers, and editors, can
> >    handle [] OWL information (as uninterpretted data). 
> 
> I'm not sure if this is a benefit.  I think of it as more of an opportunity
> for chaos.  Of course, opinions may vary.
>
> > > Drawbacks 
> > > 1/ New parsers would have to be built for OWL.
> > > 2/ An RDFS reasoner could not be considered as an incomplete OWL reasoner
> .
> > > 4/ Syntactically valid OWL would have a different meaning in RDFS.
> > 
> > How about:
> > 
> > 1/ New parsers would have to be built for OWL (but, unlike with
> >    options 2 and 4, they would be layered on RDF parsers, isolating
> >    them from RDF syntax issues and evolution.)
> 
> Again, I'm not sure how much of a benefit the parenthetical portion is.

Maybe now is a good time to state The RDF Hypothesis [1].  Informally,
it goes like this: "Wow!  RDF is so cool!  You can say anything in
RDF!  Everyone should us it!"  Cooler heads like ours might rephrase:
"Use of the RDF data model for all kinds of information (including
syntactic and meta) will result in dramatic network effects (economies
of scale), making it practical to solve many long-standing problems in
information management."

Options 1 and 3 look good to believers in The RDF Hypothesis.  Options
2 and 4 look lousy, until we bring in The Blindfold Hypothesis [2],
which says that expressions in any formal language can be read or
written as if the language were RDF (using the ontology which
implicitely or explicitely underlies the language).  With that in
mind, the WebOnt-WG using options 2 or 4 is no problem for RDF fans,
although we would like the WG to pick the URIs, etc.  If the Blindfold
Hypothesis truly holds, then options 3 and 4 can be handled
identically by appropriate software.

> Any significant change to RDF, e.g., datatyping changes, is likely to
> affect OWL. 

I was thinking more along the lines of people using other syntaxes (not
RDF/XML 1999-02-22) for RDF, like N-Triples.   (Or any language with
blindfold mappings.)

> > 4/ The syntaxes for OWL and RDFS would be disjoint (with no common
> >    sublanguage, unless some useful overlap is found), so there could
> >    be no reuse of ontology information.  (which is kind of drawback 2,
> >    again.) 
> 
> I view this as more confusing than the shorter version above.

Perhaps I misread "different meaning" as leading in the direction of
paradox again.  Any conclusions an RDFS reasoner would draw from the
triples intended for the OWL reasoner would be conclusions about the
syntactic structures of the OWL expression.    

I do see how silly it looks if you don't happen to believe the RDF
hypothesis.

> > Discussion:
> > 
> > In option 1, we tried to describe the ontological relationships using
> > RDF and found ourselves with a paradox.  With this option, instead of
> > just moving beyond RDF, we address the problem with a layer of
> > indirection: we use RDF to describe the syntactic relationships in an
> > ontology language.  It's not clear yet how much use RDF tools will be
> > with this kind of information.
> 
> This I agree with, and have added to my version of the document.

Excellent.  (Sometimes these discussions just end in confusion; this
makes me even more hopeful about this thread.)

    -- sandro

[1]  I just made this up.  I imagine members of the original RDF WG have
     been saying things like this for years, but I've never seen it
     formulated as a hypothesis.
[2]  http://www.w3.org/2001/06/blindfold/grammar

Received on Friday, 1 February 2002 17:51:55 UTC