Re: Resolution of: #rdfms-identity-anon-resources

Dan Connolly wrote:
> 
> Sergey Melnik wrote:
> >
> > Pat Hayes wrote:
> > >
> > > >It looks like it is still the case that RDF/XML syntax can only
> > > >represent a subset of valid RDF graphs,
> > >
> > > But that, I gather, is the case more generally.
> > >
> > > >in which there are no circles
> > > >that contain bNodes only. If so, this limitation needs to be stressed in
> > > >the spec(s).
> > >
> > > Agreed. There should be a whole little essay (in the primer?) about
> > > the different notations and the relationships between them.
> > >
> > > >An exception handling mechanism must be specified as well.
> > >
> > > Why? Isn't it enough that there should be round-tripping from RDF/XML
> > > -> graph -> RDF/XML?
> >
> > There should be, but probably won't be
> 
> why not? As long as the graph came from an RDF/XML document,
> I see no reason why it can't be written back out as one.

Sure, if it remains unmodified.

> > (unless the syntax subgroup would
> > attack the issue e.g. by introducing an additional attribute for
> > referencing local bNodes). Therefore, as an implementor, I'd like to
> > know what to expect when I try to serialize _x --property--> _x.
> 
> Such a graph could never come from parsing RDF/XML.
> 
> I'd suggest that if somebody built such a graph thru some
> API and then asked to serialize it as RDF/XML, you throw
> an exception.

A sentence like yours above (but of course more normative) should be
included in the spec somewhere. After all, people want (and will try) to
exchange arbitrary graphs. Do you think it is off charter to include a
new attribute like "localID" in the RDF/XML syntax to make it as
powerful as the underlying graph model?

Sergey

Received on Monday, 15 October 2001 14:54:23 UTC