Re: New syntax spec

>>>Rick Jelliffe said:
>  From: "Dave Beckett" <>
> > it used like this:
> >  [attributes]=set(idAboutAttr?, bagIdAttr?, propertyAttr*),
> > 
> > i.e the propertyAttr term can be repeated 0 or more times.  Each of
> > these propertyAttr uses can be either a typeAttr or propAttr.  I
> > think this does not restrict the use of multiple rdf:type properties
> > and/or other namespaced properties in any way.   Does that make sense?
> Specifying the schema as a grammar makes things quite difficult for
> developing a Schematron schema for RDF. (I have tried to create another one,
> b.t.w. for the new syntax: things seems a lot clearer than the old)  Because 
> one has to look at the new "grammar" and try to guess what
> the intent of the productions are supposed to be.

I am not sure what you mean by "specifying the schema as a grammar".
This is a specification of a syntax, and a grammar is the natural way
to specify such things.  The intent of the productions - i.e. what
they generate in the RDF model, is the bit yet to be filled.  We
decided to publish now to get the syntax matching ideas out there,
ask for suggestions.

> Which is a shame, because RDF clearly has the kinds of patterns that
> Schematron can express. And things like rdf:_1  (which otherwise doom RDF
> to unmaintainability) should pose no problems either.

As you can probably tell, the current section 4 is a
hand-transformation of the earlier work.  I am itching to get the
entire thing generated by some technology right now.  This document
assumes there is something we can use that can precisely record some
of the things you mention in your previous mail such as the rdf:type
/ propAttr ambiguity.

I was wondering if pattern-matching technologies such as schematron
would have a lot of problems with RDF's XML syntax since some of the
allowed things depend on the context (node or property) which can't
be determined without considering the containing XML elements etc.


Received on Thursday, 13 September 2001 09:11:02 UTC