On Tue, 19 Mar 1996, Paul Prescod wrote:
> In other words, why require extra validation smarts at the user-validator
> level instead of at the DTD-validator level? Presumably DTDs are validated
> much less frequently than are documents.
These proposals break because they demand that there be a DTD which
which describes all possible combinations as opposed to having two
DTDs which are applied conditionally. THe CPP and MARKED section
approachs don't require:
1. COntent providers to be DTD lawyers
2. Every UA to be capable of groking every arbitrary DTD and mapping
it against the UA's capabilities.
In addition, there are UA characteristics which are not ever likely to
be represented in a DTD. For example, hardware characteristics of the
UA's execution environment.
THe advantage of conditional markup is it is VERY flexible and has been
proven over time in many other contexts of being able to deal with
issues not identified when the markup was designed.
THe <ALT> proposals, the <NO*> proposals etc. all become rather quickly
embroiled in discussion of how to order things what the content model
might have to be, what the restrictions are on extensions to make it
We need and will continue to need a solution which transcends those
concerns. Lets create an extensiblity environment where we can devote
our energies to design of the functionality not how to retrofit it
to the existing world.
Don't presume that a situation has a single alternative. Don't require
any understanding, including access to new DTDs, on the part of
already deployed UAs for the new features/markup to be tolerated