On Feb 25, 2006, at 01:16, Ian Hickson wrote:
For the kind of detailed conformance checking HTML5 requires,
you'll be wanting to use your own explicit hard-coded
implementation, not
one or more schemas. IMHO.
Is there any particular reason why you think one wouldn't want to use
schemas?
The downside I see is that error messages will be less precise than
what carefully crafted custom code would allow. However, I believe a
schema-based implementation to be easier to write, more maintainable,
more portable and better customizable (portable and customizable e.g.
for inclusion in a CMS that wants more restrictions on input).
My current plan is
1) to use RELAX NG for everything it is convenient for (element
nesting, attribute occurrence, attribute datatypes through datatype
libraries)
2) to use the XSD datatype library within RELAX NG for everything it
is good for
3) to define (in English) and implement (in Java) a datatype library
for stuff that the XSD library is not good for
4) to use Schematron for stuff that Schematron can cover but the
above can't conveniently cover
5) to use custom Java code if there is still something that is not
covered.
--
Henri Sivonen
[EMAIL PROTECTED]
http://hsivonen.iki.fi/
