Re: proposed interim response for ``Comments from RDFCore on S+AS''

* Peter F. Patel-Schneider <pfps@research.bell-labs.com> [2003-05-09 14:52-0400]
[...]
> > Specifically we request, that in Owl DL and Owl lite:
> > 
> >    a) that a b-node representing an individual may be the object of more
> > than one triple
> >    b) that cycles of b-nodes representing individuals be allowed.
> 
> We note that this is a restriction that until recently has also been in
> RDF.  It is thus hard for us to see how this can break any significant
> number of deployed uses of RDF.

The only restrictions were accidental ones imposed by RDF's primary XML syntax. 

RDF since the 1999 REC has permitted graphs with b-nodes arranged in ways that 
OWL DL/Lite now disallows. As you point out, exchanging these graphs in XML 
prior to rdf:nodeID was tricky, but there are plenty of deployed RDF systems 
out there where APIs (SiRPAC, Mozilla, Jena, Redland, ...) or graphical UI
(RDFAuthor, IsaViz) or alternate syntaxes (N3, XSLT transforms) are used to 
create and exchange RDF graphs. 

B-nodes are a popular construct, in 
large part because they allow sidestepping of angels-on-pinhead arguments 
like "what is the URI for a person? a company? an event...?". B-nodes are 
significantly deployed, and adjusting those deployed apps to fit with DL/Lite 
would be a major undertaking. For example, consider the RDF datasource system 
in Mozilla. Currently you can assert new statements into any of its graphs 
(datasources), ie. bookmarks, in-memory graphs, etc. They _could_ adjust the 
RDF API so as to catch any attempting assertion of a triple, check it wasn't 
made against a b-node, and throw an exception. Or silently fail? Or mark the 
graph OWL-Full? Many Mozilla-based applications use their RDF API and were 
constructed without concern for this constraint. I believe the same to be 
true of Jena-based and redland-based apps...

Dan

Received on Friday, 9 May 2003 22:49:03 UTC