>I think your model of "compound document" is too simple. You seem
to model it as a set of self-contained fragments enclosed in a
container.
Most current proposals for using namespaces take just that approach.
Namespaces were initially conceived as a way of allowing controlled sets of
metadata to be added to files. Their extension to a general purpose
mechanism, while useful, should not blinker us from the real user
requirement, which is to take data generated by other applications and add
it to a file that has previously been validated using a specific DTD.
>I expect to see much more complex documents: a Chemical ML description
of a molecule, with (a) embedded HTML English-language documentation,
which itself contains several inclusions of MathML, and (b) a SMIL
document that marshals a non-XML movie of the molecule. How would
you validate just the HTML part, given that the HTML DTD does not
understand MathML?
I would not try to validate the whole thing in one go. I would create the
MathML within a Maths generator module that could only generate validated
MathML. I would validate any HTML I needed to embedded within the document
with an HTML editor which had enough sense to ignore any element that had a
namespace other than HTML specified, together with any nested elements that
were not declared to be HTML compliant. I would edit the Chemical ML
material in an SGML/XML editor that was set-up to validate material in that
DTD, but ignore any element (and its contents) that had a namespace
specification for another DTD.
Martin Bryan