RE: Namespaces, schemas, Simon's filters.

[Jeff Lowery <jlowery@scenicsoft.com>]

> I see nothing wrong with a rule that says if you're going to 
> give something
> a name, give a unique name. There's not way to enforce that, 
> of course, but
> we're dealing with it for namespace identifiers already.

Er, should have said "disambiguatable name", not "unique name". This would
include non-unique local names that are disambiguated by the context of
being in a parent element, and so on up the chain, until an ancestor element
is found that is global w.r.t. a namespace.

Thus local names are associated with a namespace (and I'll take any
definition of "associated" you care to give), and any context that the
namespace might connote would be inferrable.

Again, I would like to know why it was decided that namespaces be
represented as attributes rather than parent elements, since it seems
elements (superficially at least) are a more natural representation.
Understanding breeds contentment, right?