[
Lists Home |
Date Index |
Thread Index
]
Tim Bray wrote:
> I mostly agree. You've argued that doing the following
>
> <x:y xmlns:x="http://example.com/"; x:a="1" a="2" />
>
> while legal per the ns rec, is idiotic and shouldn't be done. Agreed.
> You've further argued that it was a design error that the ns rec allws
> this. Agreed.
Cool. I didn't think we were that far apart on that aspect.
> Where we may disagree is, if you're making an API, you'd better not
> report the final attribute above (to use JJC's notation) as
> {http://example.com}a, whether or not the first attribute is there or
> not. For better or worse, at the API level, the final attribute there
> has a local part of "a" and no namespace name, but is attached to an
> element with a namespace name. -Tim
On this, I agree that generic APIs (SAX, DOM, etc.) and XML processing
tools (XSLT) would be wise to follow the approach you outline. I think
some kind of principle of least surprise makes sense there, at least for
now. (I'd want that re-examined if XML 2.0 ever came to light.)
Within specific (software) applications, I'd leave that up to the
application, but given the kinds of mapping that go on above the parser
level anyway, this shouldn't be exactly shocking.
--
Simon St.Laurent
Ring around the content, a pocket full of brackets
Errors, errors, all fall down!
http://simonstl.com
|
