Re: XHTML m12n XSD

[Rick Jelliffe <ricko@allette.com.au>]

From: Arnold, Curt <Curt.Arnold@hyprotech.com>


> I'd be curious to know what you liked about "redefine".  It would seem to
me that you could accomplish typically uses of redefine with XSLT during
schema authoring time and there is no need to burden
> the validator with the complexity of fetching all the previous versions or
variants of a schema and performing an invariant merging.

Yes, <redefine> <include> and <import> are all just shorthands to make
declarations easier.  But the point of m12n is to make a set of abstractions
which concveniently group XHTML declarations together:  in m12n you use
naming conventions on top of PEs to represent the abstract m12n, in XML
Schemas we have some more specific inclusion constructs than PEs which allow
a little higher-level modeling of the m12n abstract model.

> It is something that is essential to XHTML m12n or is it just a
convienience?

I think it is essential to XHTML m12n, but obviously not essential to XHTML.

There are many equivalent ways that the same schema could be structured: for
example, we could do away with <redefine> ( in the unextended XHTML schemas)
by having two sub-schemas for each module, one giving the effect of the
INCLUDED module, and one providing the effect of the IGNORED module (i.e.,
one would just change the URL to INCLUDE or IGNORE).   That would be nice,
but I just couldn't reconcile it with the m12n anstract model, which seems
to be based on a module being included or not:  if the XHTML M12n WG decides
this is not intrinsic to their idea of M12n we could use this dual-schema
approach (which is partly needed because there is no equivalent to <!ENTITY
% x ""> in XML Schemas, I think)

The underlying issue is this: is it possible to enumerate all the possible
uses of PEs and provide direct high-level equivalents of them (in present or
future  XML Schemas?)   I think it is certainly worthwhile doing this for
one level above PEs, which is what <import>. <include>, <redefine>,
substitution groups and type derivation try to do: but no-one has figured
out every possible construct that can be built with PEs -- things like m12n
are useful but certainly not exhaustive for what can be built.

XML schemas does not have an equivalent to IGNORE/INCLUDE, and Dave Peterson
has long called for the introduction of boolean expressions in
INCLUDE/IGNORE sections.  I think it comes down to whether people expect
schemas to maintained in entity managers or in proprietary systems.

Cheers
Rick Jelliffe