Re: some questions on RefactoringRDF draft

Aaron Swartz wrote:

[...]
> > Does that imply an RDF processor MUST attempt to create tripels
> > from any wellformed XML??
> > How does an application determine WHERE to start processing for RDF??
> > Can it just make arbitrary guesses with a given XML document??
> 
> This is a matter of some discussion among the working-group. It
> seems the best way to do this is to look for the RDF namespace,
> or have the user of your parser make it clear where the RDF
> starts. However, the issue isn't closed yet.

Whilst we have not defined a processing model for RDF, I think the
answer to the question asked is that NO an RDF processor
is not expected to process arbritary XML.  It is only expected to
process RDF/XML.

There is some discussion in the WG about whether the <rdf:RDF></rdf:RDF>
brackets are necessary to mark RDF XML, but I think that's a different
question.


> 
> > typedNodes MUST be prefixed by now.
> >
> > Does that imply RDF will NOT allow any kind of namespace defaults??
> 
> Hmm, perhaps the terminology could be clear here. Namespace
> defaults are OK, as long as they are clearly defined. That is,
> you need to have an xmlns="..." definition in your XML. Then the
> nodes are prefixed in the Infoset, even if they don't have a
> prefix in the XML. (This is rather confusing...)

We are aiming to switch technology and terminology here.  We are
moving to define RDF/XML in infoset terms, not in terms of a
bnf grammar on text.

What we have said is that typedNodes are required to be
namespace *qualified* (if we have goofed and said prefixed somewhere,
please tell me where and we'll clean it up).  A typed node element name
can be namespace qualified either by having a specific prefix or
using the default namespace mechanism.

Where did you see a suggestion that typed nodes needed to be prefixed?
Maybe we can make this more clear there.


> 
> > ALL attributes MUST be prefixed
> >
> > Does that imply an RDF processor MUST attempt to map all
> > wellformed XML but will fail almost certainly??
> >
> > Maybe it's not ALL attributes, but just those, which are
> > supposed to come from RDF??
> 
> Yes, a parser is only supposed to deal with the RDF portions of
> a document.

There is no requirement that an RDF processor process all of XML,
only the RDF portions.  But within RDF/XML all attributes must
be prefixed.  Actually there is at least one exception to that
rule, xmlns, but this is a special case.

I have been thinking in terms of eith/or scenarios - in which some
XML is either RDF or its not.  If it is then it must conform to
all the rules.

Are you suggesting that there are scenarios we should consider where
you would like an RDF processor to extract what it can from the bits
of the XML that match the RDF and ignore the rest?  If so, a real
use case would be very helpful to us.

> 
> > RDF M&S requires ALT containers non-empty. This restriction
> > seems to be removed - intentionally??
> 
> Hmm, I was not aware of this issue. Brian, can you perhaps shed
> some light here?

So far as I am aware this restriction has not been removed, though
we have not finished discussing containers yet.  What was it that
suggested to you this restriction is gone?

Brian

Received on Tuesday, 11 September 2001 04:52:46 UTC