> > And I'm not talking about exotic, rare and proprietary schemas. To begin
> > with, some examples are all XHTML documents that have modules in other
> > namespaces, like RDDL and WAP 2.0. There are DTDs and schemas for those
> > language, but RDDL is not fit to handle them.
>
> As I said in my previous message, I don't see this as a problem with RDDL.

As I've said three of four times now, show me the algorithm that will find
the URL of the RDDL document for this document : http://www.rddl.org/. This
is a FONDAMENTAL problem with RDDL.

> Instead I think it suggests that our approach to writing *schemas* isn't
> flexible enough to deal with documents containing an arbitrary mix of
> namespaces. I should state up front I don't have any answers here,
> I'm just interested in a general discussion.
>
> The majority of schemas that I've seen assume a fixed set of
> elements. These elements may come from zero, one or more namespaces.
> The schema is "closed". They're designed to validate a particular
> class of documents.
>
> Some schemas are "open", i.e. they allow "unknown" elements to be
> used in the document, and these usually in fairly fixed places (cf: XSD
> ANY, XHTML DTD Modularization). However these schemas still seem
> to be designed to validate _documents_. They validate the document,
> and ignore sections of it, or as with modularization defer to other
> schema/dtd modules.
>
> Yet the scenario you're discussing is one which seems like it could
> become increasingly common: we have a mixed namespace document
> for which there is no schema. You're asking, how can I validate
> these documents? Is there a heuristic for combining together several
> schemas to achieve this goal?

In fact there are schemas for RDDL and WAP 2.0, DTD are provided for both
examples. Nothing prevents a schema from manipulating names from different
namespaces. However, the scenario I'm discussing is a scenario where the
document has not a single namespace, breaking the false but too often
believed assumption that a namespace name is equivalent to a DOCTYPE
specification.

Regarding "open" schema, that is to say schema that are meant to be extended
or inherited, there is indeed a problem. I think namespaces are part of the
solution, but are in no way THE solution.

> To do this you need to define schemas not only to be open, but also
> to be easily fragmented so that portions of it can be applied. I can
> imagine doing this with a schematron schema (only apply certain
rules/patterns),
> but not with a DTD. I also assume there's a way to do this with RELAX NG
> and XSD. You then need to apply these fragments to the document to
> validate it.

Yeah, I think that in opened schemas, partial validation of "patterns" are
required, instead of global validation of whole documents. This can be done
with the current schema languages, though, even DTDs. I tried to write down
some thoughts about it there :

http://nicolas.lehuen.com/Articles/Programming/Ideas/fog0000000033.html

>
> I don't see schemas being written with this use in mind, nor do I see
> validators that allow this flexible application of schemas.
>
> I may be showing a lack of understanding here, I don't mind
> looking a fool :) Does anyone else see this as an issue, or is it
> not a problem? Am I misunderstanding something?
>
> RDDL only enters this picture as a way to associate a schema, or
> fragment thereof with a Namespace URL. Doing something with those
> fragments, assuming they're available is something for the validator.

Schemas can be associated to namespace (one or many namespaces), but the
contrary is, IMHO, a bad idea. I am sorry, but I don't see how to solve this
problem without relying on the concept of document or element type. I'll try
to expand on this in the article mentionned above.

Regards,
Nicolas

> Cheers,
>
> L.
>
> --
> Leigh Dodds, Research Group, Ingenta | "Pluralitas non est ponenda
> http://weblogs.userland.com/eclectic |    sine necessitate"
> http://www.xml.com/pub/xmldeviant    |     -- William of Ockham
>
>