[
Lists Home |
Date Index |
Thread Index
]
Norman Gray wrote:
> Indeed, indeed. All the examples use attributes in no namespace. If
> you create a stylesheet with attributes in the XSLT namespace, it will
> not work. There are _no_ attributes in the XSLT namespace, it seems.
> [...]
> The original point was to suggest a case where it would make a difference
> if one were to change the namespace spec so that no-prefix attributes
> inherited the namespace of their enclosing element. If this change were
> made, XSLT stylesheets would probably break. That's all.
Possibly -- but the problems would be with result elements,
*not* with elements in the XSLT namespace.
The XSLT specification would break, since there would then be
no way to add non-namespaced attributes to namespaced elements
as is currently required. So the XSLT REC would have to be amended
to account for the new XMLNS rules, replacing unqualified attributes
with qualified ones in the XSLT namespace.
New [XSLT] plus old [XMLNS] would break stylesheets, and old [XSLT]
plus new [XMLNS] would be impossible, but as long as implementations
change to support the new versions of both at the same time,
existing stylesheets would remain legal (except for pathological
cases like "<my:foo my:bar='1' bar='2'/>").
The *output* of those stylesheets, however, would change:
previously unqualified attributes on namespaced elements
would acquire a namespace name. Of course once the vocabulary
of the result document has *also* been updated to reflect
the new [XMLNS] rules, all is well again.
I suspect that most namespace-aware vocabularies could be adapted
to the new XMLNS rules, with minimal impact on existing documents
and processors. But it would have to happen all at the same time;
the old rules are incompatible with the new ones.
--Joe English
jenglish@flightlab.com
|
