Re: [xml-dev] Global or shared namespace?

[David Brownell <david-b@pacbell.net>]

> > > 3) How does DOM L2 not conform to the NS Rec?
> > 
> > It puts namespace declaration attributes into a specific
> > namespace, contradicting a requirement (in the text
> > quoted above) in the namespace REC.
> 
> While this is true, the DOM conforms to the Infoset spec [1], and that
> behavior was agreed upon by the W3C XML Core WG, as stated at [2]. 

Since DOM L2 predates the infoset REC, that's sort
of a bizarre way of looking at things.  It's more like
"Infoset conforms to DOM", or "both conform to [2]".

And [2] is rather curious ... it's not a REC, was published
a week after the last erratum to the REC, "overlooked" the
W3C processes that exist to prevent incompatible changes
to a REC, yet is being treated as normative, and:

    - It's a namespace URI that actually points
      to something -- completely unexpected!!

    - The explanation seems dubious.  It says
      "was not associated", implying oversight
      where there was instead explicit prohibition.

      And it seems to overlook the "not in any
      namespace" case, at least for attributes,
      where the namespace REC talks about
      a "Per-Element-Type Partition" (in that
      strange non-normative appendix).

I think we can all agree we've noticed a couple
more cases where "Namespaces in XML"
create more problems than necessary.

Though in this case it's the process of [2],
not the REC ... the folk writing the REC, to
my strong recollection, explicitly rejected that
approach.  Ditto its extension to letting any
prefix in a namespace like [2] be another
way to declare a namespace.


> The current situation where various W3C
> specs disagree on this is unfortunate

I'd put it more strongly, but maybe that's
just me.  [2] was a process failure, and
it's been unresolved for almost two years.

- Dave


> > [2] http://www.w3.org/2000/xmlns/