Lists Home |
Date Index |
"Simon St.Laurent" wrote:
> While I agree with you that there really is no "ideal text", I have a
> very hard time finding that a justification for the entirely new burdens
> that Namespaces in XML created.
> It makes me laugh and cry that I should have to read through all of the
> attributes of a given element just to figure out what the
> properly-constructed name of that element is.
back in 1998 (that is, with the appearance of draft 2), that requirement struck me at first as an inadequately motivated degree of flexibility. upon reflection,
i realized that a processor did not have to do anything except intern any NSAttNames and assert the namespace bindings immediately. by that point, the cl-xml
had already been interning names for a year, so _nothing_ else changed. that was two lines of code. one to intern the name, the second to bind the prefix to the
namespace name. in order to accommodate scoped namespaces i had to ensure that the bindings had dynamic rather than indefinite extent. that was another line of
code. more accurately an additional &aux binding. in a language without primitives to express dynamic bindings, it might be a bit more work.
relative to the size of the parser, that is not an excessive burden.
as i noted above, i'd already, by that point, been coding applications with respect to interned universal names. _none_ of that code changed.
yes, it does mean that an application can act on names directly off the wire only if it can require that namespaces are declared in an ancestor element. is that
really a problem relative to the other constraints on such an application?
> The complicating factors
> that XML 1.0 itself introduced - default attributes and entities - at
> least didn't involve that kind of jumping back and forth.
> Namespace semantics are a bad joke.
"namespaces in xml" has a very clear semantics. having to do with the universal name denoted by a given (prefix x local-part) pair.
> Let's stop pretending they ever
> existed, since no one has ever been able to explain what they are.
i've never "pretended" anything else.
> Labels without built-in semantics appears to be as far as we've ever
> gotten, and that's hardly worth the representational hassle.
my post was not cencerned with these issues. please do not confuse the two.