Re: URIs for "xml" and "xmlns" prefixes?

David Brownell wrote:
> 
> James Clark wrote:
> >
> > David Brownell wrote:
> > >
> > > The namespace spec states that there are two "implicitly declared"
> > > namespace prefixes, "xml" and "xmlns".  But it doesn't give their
> > > associated URIs, as needed for uniformity within implementations.
> >
> > I don't see a need for a URI for "xmlns".  Names with the "xmlns"
> > prefix  are never seen by applications.
> 
> That's not in the current spec.  How will applications know where the
> namespaces are declared, if not through this (IMHO) natural model?
> Or you assuming that preserving the declaration structure of source
> data is not a goal?

You were thinking that you would expose the declaration structure by
passing through the expanded xmlns attributes? I hadn't thought of doing
it like that.  There are other ways to do it (see XT for one example).

> I note that the XSL draft includes a requirement that applications be
> able to see the list of namespace prefixes that are "in scope".  (2.4.2)
> Looks like some API-like requirements are bubbling up; I don't think they
> would naturally belong in this document.

I think some kinds of application will need this (though they'll
probably be in a minority).

> > We know we need a URI for "xml".
> 
> Good ...
> 
> > The intention on "xmlns" is that's it's a keyword.
> 
> That might simplify some things.  Like supporting default namespace
> declarations.
> 
> But -- just "xmlns", and not "xml"?  So it'd be allowable to say
> "xmlns:xml='http://www.example.com'" and thus change the meaning
> of an "xml:space" or "xml:lang" attribute?

I don't think I would want to allow that. It's not obvious exactly what
to do here, so I think it's best to keep it simple (restrictive) for
now, and maybe generalize later if the simple approach turns out to be
problematic.

James