Re: forest grammar/tree regular expression for RDF (fwd)

On Sun, 24 Jun 2001, Dan Connolly wrote:

(on Jonathan Borden's RELAXNG work...)

(msg trimmed to focus on rdf:Description question)

> > I have specified this as a RELAXNG schema for RDF
> > http://www.openhealth.org/RDF/RDF1.rng in terms of solidifying the RDF XML
> > syntax under the issue:
> > http://www.w3.org/2000/03/rdf-tracking/#rdfms-formal-grammar
> 
> Very nice!
> 
> It has a few special cases that I don't think are necessary
> (e.g. rdf:value is just another propertyElement/propertyAttribute,
> and rdf:Description is just a typedNode) but other than
> that, it seems to be quite a compact and precise description of
> the RDF syntax.
> 

[...]

> Production 4 is ambiguous, no?
> 
> 4.type description = 
>              rdf:Description[
>                      idAboutAttr?,
>                      bagIdAttr?,
>                      propAttr?,
>                      propertyElt*
>              ] |
>              typedNode
> 
> <rdf:Description/> matches both alternatives, no?
> 
> As I say, I don't see any need to special-case
> rdf:Description in the grammar.


If we take this reading of the syntax, then the presence of
rdf:Description asserts an rdf:type relationship between the described
resource and an rdfs:Class called rdf:Description.

I've seen nothing in RDF elsewhere to support the claim that RDF defines
such a class; M+S is pretty clear that the rdf:Description construct is
pure encoding syntax.

If we were to decide that such a (goofily named) class exists, would it
be something like a subclass of rdfs:Resource?  Your proposal seems to
make the rules for our XML encoding syntax simpler at the cost of making
the resulting structures more complex. 

Dan