[
Lists Home |
Date Index |
Thread Index
]
Elliotte Rusty Harold wrote:
>
> It probably looks like what was used in an earlier draft of the
> Namespaces spec, where all declaration were made with processing
> instructions in the prolog:
...
>
> I suspect an irrational aversion to processing instructions killed this
> sensible scheme. This would not have fixed the issues with unresolvable
> URLs or attributes in no namespace, but it would have neatly eliminated
> all scoping issues.
This wouldn't have solved all the problems. Any XML API is going to
expose PIs so people could have hung themselves manipulating them.
But the real argument that carried the day was the hypothesis that if
you're streaming out XML, scoped namespaces allow you to stick something
in the middle that you didn't know about when you were doing the root.
I'm not sure how often this application scenario has arisen in practice,
but it was a pretty convincing argument at the time. -Tim
|