This was discussed by the WG and eventually voted down; the advantages
of the attribute approach, combined with the desire to avoid having
multiple ways to do the same thing, led to the PI approach being
amputated.
>without such a mechanism, the dtd's denotation is undefined.
I don't understand what you mean by "denotation". In general, the
usefulness of your postings would be improved by more attention
to the use of terminology, and where you need to use a new term
(e.g. "extent" in your previous posting and "denotation" here) you
should provide a definition for it.
>if, on the other hand, such a mechanism is provided, then validation is
>possible on the basis of namespace-aware 1.0 conformant dtd's.
Only trivially so. As I have pointed out, the hard part is not matching
up the names, it's making compound DTDs. I have also pointed out how
a mechanical instance/DTD rewrite could address some of the issues
of namespace-aware validation. I have not suggested a solutoin for
how to solve the *hard* problem, namely how to go about making a
compound DTD, and would like to hear input on that. -Tim