- From: Masayasu Ishikawa <mimasa@w3.org>
- Date: Tue, 13 Aug 2002 22:46:33 +0900 (JST)
- To: www-html@w3.org
Lachlan Cannon <luminosity@members.evolt.org> wrote:
> > Why stick with the DTD, when it's so apparent that it cannot describe
> > the language correctly? XML Schema REC provided a non-normative DTD
> > for Schemas, and I could live with that approach. People could use
> > it at their own risk, but I'm not tempted to requiring validation
> > against that DTD to user agents.
>
> Well yes, that's more or less what I was getting at. Out of interest
> does XHTML 1.x require validation?
- *documents* are required to be valid
- user agents are NOT required to perform validation, they only check
well-formedness and throw away not well-formed document
Validation is the task of the sender of resources, not the receiver.
> If so, then current browsers don't
> even care (apart from Moz with the right doctype as far as I'm aware),
Mozilla would switch different "mode", but that's different from
validation. Mozilla recognizes a small set of DOCTYPEs and read
local copy of DTDs, but those local DTDs only include declarations
for so-called character entities and don't include element and attribute
declarations, so doesn't help validation.
> You could produce a schema of course, and then produce a non-normative
> DTD that expresses everything the same but in terms of default
> namespaces.
We could provide a non-normative DTD, but if it's non-normative,
don't assume you can omit the namespace declarations from your
document.
> I think it's definitely a good idea to at least publish this
> non normative DTd however, because they're a hell of a lot easier to
> read a quick fragment and see how something works (cf the XLinks spec).
Well, modular DTDs are not so easy to read, I would say. I personally
find RELAX NG a lot easier to understand, in particular when a schema
is modular.
> > Man, I wrote that DTD driver and I know how painful it is than
> > anybody else. It's too tricky that sane people should never bother,
> > it very clearly illustrates how inappropriate DTD is for describing
> > a multi-namespace document type. I said, if we are ever able to
> > write a DTD for XHTML 2, most probably it's even more complex than that.
>
> Well, I'd trust your word on it then (Not that I don't love the way that
> thing works.. I'm using it myself. Very nice).
It's good to know someone tried it, thanks. The most unfortunate
restriction in the XHTML+MathML+SVG DTD is that authors cannot
use namespace defaulting for XHTML and SVG at the same time due to
name collision. When an author tries to include an SVG fragment
inside XHTML, the most natural usage would be just declare
the default namespace for that SVG fragment and won't prefix SVG.
Vast majority of SVG document out there are not prefixed, and
people would like to just copy and paste them inside XHTML.
It's perfectly a legitimate use case and should just work that way,
yet DTD cannot allow such usage.
cf. http://www.w3.org/TR/XHTMLplusMathMLplusSVG/#bug-name-collision
If we include some portion of SVG in XHTML 2.0, we'll face the same
problem so long as we use DTD.
> It makes me wonder...
> should the W3C rewrite the XML DTD wrt namespaces to make it easier to
> do this stuff? Once again, I personally think DTDs are much more
> readable & intuitive.
That's what ISO/IEC 19757 Part 9 is trying to cope with, and while
I cannot speak for the W3C's plan for the XML DTD, my guess is very
skeptical at best.
> In any case, I think that a non normative DTD with default namespaces is
> definitely a must have appendix. The actual validation should of course
> be written in schema if that's more appropriate and easier.
Once again, whether we provide a non-normative DTD or not, don't
assume you can omit the namespace declarations to be conformant.
> After all
> that gives the HTML Working group more time to work on other aspects of
> the spec, which can only be a good thing, ne?
Well, it comes down to when *I* do that (I'm the editor of DTD part),
and whether people can convince me that it's a good thing ...
Regards,
--
Masayasu Ishikawa / mimasa@w3.org
W3C - World Wide Web Consortium
Received on Tuesday, 13 August 2002 09:46:42 UTC