Lists Home |
Date Index |
- From: Paul Tchistopolskii <firstname.lastname@example.org>
- To: email@example.com
- Date: Fri, 29 Dec 2000 14:46:24 -0800
----- Original Message -----
From: Jonathan Borden <firstname.lastname@example.org>
> The point has been raised that a namespace URI identifies a single
> schema (forget content negotiation for the moment). My point is that
> doctypedecls also identify a single schema, hence this problem is not new.
Not really. There are some differences.
1. doctypedecls are pointing to DTDs. They're pointing
to some standard thing ( DTD is standard. Useless, but standard).
Namespaces are poiting to nowhere.
> If this argument is used to claim 'namespaces are broken' as I've heard,
> then it equally claims 'XML 1.0 is broken'. Of course nothing is perfect, so
> everything is broken.
DTD layer in XML is of course broken ( because it does not survive in the
real life for another W3C standards, like XSL FO, for example ), but let us
forget about it.
My point is that namespaces ( pointing to nowhere ) are broken in
2. Also, DTD layer could be ignored. I mean you can just not use DTD layer -
and that's it.
*Namespaces* are much harder to ignore. I still think the purpose
of namespaces was/is to assign a unique context, not to fetch the URLs.
Not that I enjoy bashing namespaces ( even I still think that SGML people
were right and resolving form context is better than resolving from hidden
'second-name' ), but I think that situation with namespaces is different
than situation with DTDs.
3. Let's think a bit about the DTD layer. We have DTD. We have a
doctypedecl. doctypedecl is like a PI. It says : "validate with this DTD".
Current semantics of namespace 'PI' is different. I'd say that the semantics
of namespace PI is "handle prefixes/hidden lastnames". Adding *more* semantics
to 'namespace PI' is just wrong, I think. It'll produce a monster 'PI' with
2 orthogonal semantics attached. To me it smells like a suspicious step.
Again - I'm not sure I can say is it good or not ( in my universe there is
no namespaces, because I think that resolution from the context is
the way to go ) - but the point of this letter is explain what
is different between namespaces and doctypedecls from my point of view.