What the namespace spec should say (was: Re: Irony heaped on irony)

John Aldridge wrote:

> At 08:48 26/05/00 +0100, David Brownell wrote:
> >The "XML Namespaces" specification is quite clear that the
> >purpose of a namespace "Universal Resource IDENTIFIER" is
> >identification, not location.
>
> But this whole debate is not about what "XML Namespaces" says (that's
> largely uncontentious), but about whether it should be amended to say
> something else.
>
> I think that TimBL and DanC believe that a common (if not the only)
> mechanism for finding metadata about a namespace will be to dereference the
> namespace URI.  (If I'm misrepresenting them, I'm sorry, and I hope they'll
> clarify their position).
>
> Others on this list (including myself) disagree.
>
> These two positions place rather different requirements on the behaviour of
> the namespace URI.  Specifically, they lead to opposite conclusions on
> whether the NSURI should be absolutised before comparison.

I originally felt very strongly that the position you describe as that of
TimBL and DanC was the right one, perhaps because I found it incredible that
the behavior would be anything else.   I still feel it's essential to have
some standardized way of associating metadata with namespaces.  But the
discussion has shown there's another way to do that: create new attributes for
that purpose.

[Side note: the specification of such attributes could be tricky, because
attributes are normally treated as a set, not a sequence.  The xmlns-metadata
attribute would have to specify both the namespace prefix and the URI for the
metadata (and maybe its intended interpretation also).   Ordinarily a single
attribute specifies a single property.  This looks to me like a difficult
problem but not an insoluble one.]

Paul Abrahams

Received on Friday, 26 May 2000 12:17:32 UTC