I certainly hope that HTML's evolution has progressed beyond
that stage, though!
This is what I called "working against SGML instead of with it"
in a recent message to html-wg [1]. Starting with a syntax
and trying to reverse-engineer a DTD that accepts that syntax
is a path to frustration and despair. But now that HTML 2.0
more or less successfully describes the current state of
affairs, there is no reason at all to continue doing things this way.
If future extensions are designed, defined, and described in
terms of SGML from the beginning, the formalism will prove
to be more beneficial than burdensome.
If that's not to be the case, and designers continue to work
backwards from a preconceived syntax to a post-hoc formal definition,
I strongly urge that SGML be abandoned altogether before any more
damage is done. HTML won't gain anything by using it, and will
give SGML a bad name to boot.
--Joe English
[1] I can't get to the archives at www.acl.lanl.gov at the moment; see
<URL:http://www.crl.com/~jenglish/sgml/work-with-it>