Re: A radical finding on Using Qualified Names (QNames) as Identifiers in Content

an architectural finding which stipulated dynamic extent for 
prefix/namespace-name bindings would be a good thing.

it is more a "strict" view of their use than it is a "radical" one.

among other things, it would permit future versions of the xml information set 
to deprecate in-scope namespaces.

...

Norman Walsh wrote:
> 
> 
> Having considered Rick's message, I can see the following radical alternative.
> (I'm not sure I support it, I'm just floating it for comment.)
> 
> 1. Stipulate that xmlns declarations are [for bindings needed by
>    the parser *only*]. In other words, for qualified element and attribute 
> names.
> 
>    This gives the XML parser complete freedom to discard any namespace
>    binding that it does not need.

i would understand the first sentence to imply that, once element and attribute 
names have been resolved, a processor would be
licensed to disregard *all* bindings, if not to discard them.

> 
> 2. Since all qnames in attribute values and element content exist only
>    for some application to process (the parser can't see them),
>    stipulate that the application must provide some other mechanism
>    for associating prefixes and namespace names.
> 
>    This puts those extra bindings into the content of the XML document
>    (or entirely out of band) in ways that no processor would consider
>    discarding.
>