Re: Using RELAX NG in XHTML 2.0

> Daniel Glazman <> wrote:

> > Sorry but I don't see browsers implement that any time
> > soon.

Karl Ove Hufthammer <> wrote:

> Well, I used to think validating was an important feature for
> browsers, but I'm not so sure anymore. Validating is really a
> *tool* for *authors* to ensure that they produce high-quality
> documents / technically 'correct' documents / documents that
> 'make sense'. This doesn't *need* to be implemented in browsers.

Exactly.  XHTML2 *documents* need to be valid to be conforming,
but XHTML2 user agents won't be required to perform validation
at run time.  In that sense browsers are not required to read
external DTD subset nor XML Schemas nor RELAX NG schemas.

> > Third, knowing as soon as possible if the modularization of XHTML 2.0 
> > can or cannot be described by a DTD seems to me of *critical importance* 
> > for browser vendors.

Personally I've said many times on this mailing list that DTD cannot
fully describe the syntactic constraints of XHTML 2.0.  There are
at least three elements (e.g. 'object') whose content model cannot be
expressed by DTD but can be expressed by RELAX NG.  Handling namespaces
in DTD is hopeless.  XForms makes it highly impractical to even think
about DTD validation.  Do I need to go on?

Having said that, the WG would try to approximate intended constraints
to the extent possible by DTD if people still find it useful, but
I'm still not convinced that requiring DTD validity is a good idea.

> > Again, browsers use DTD or Schemas, not RELAX NG.
> Not any that I know of. Which ones are you thinking of?

Some browsers may read external DTD subset, but I'm curious to know
which browser uses XML Schemas for rendering a document.  Do you
expect browsers work on Post-Schema Validation Infoset (PSVI)?  If
we somehow get rid of character entities and solve ID issue (see
the TAG issue xmlIDSemantics-32 [8]) without relying on DTD, for
what purpose browsers use DTD at run time?  To get default values?


Masayasu Ishikawa /
W3C - World Wide Web Consortium

Received on Wednesday, 14 May 2003 06:42:16 UTC