Pete Cordell wrote:
> The namespace problems are often mentioned. Are there any pointers to
> how, with the benefit of hindsight / no baggage, XML 1.0 + namespaces
> should have been done?
URIs are the cause of much of the theory pain - specifying a smaller
mechanism without the heavy baggage (and promises) that URIs carry would
have been helpful.
I'm not thrilled about the declaration syntax, but that's one case where
I haven't really found a better option.
In any case, my proposal is _just_ subsetting, not redesign. Redesign
is, I think anyway, better done through developing a new spec, under a
different name, etc. (See JSON for one example.)
> (Certainly from a databinding point of view it would be nice if the
> worst-case number of characters you had to look-ahead to work out the
> namespace of an element (or attribute) could be predictable.)
Personally, I'd call that a severe case of over-optimization. If people
are sending you documents with namespace prefixes that are more than 20
characters long, something wacky has gone wrong.
> P.S. If we're looking for features, how about an xml:type attribute
> rather than having to use xsi:type?
Wow - now that's precisely the kind of potential disaster I'm hoping to
avoid by focusing on the layer _below_ schemas. Let's _not_ look for
new features.
Thanks,
Simon St.Laurent
Retired XML troublemaker
http://simonstl.com/
