I agree with Nathan that HTML cannot be all things to all
people for semantic encoding. The AAP DTD tried to do that
and turned into a monster. I would hate to see WWW get
bogged down with a huge DTD that intimidates new authors
and makes performance a problem.
SCO does not plan to directly author in HTML. We view it as
a presentation form, and will code our source files with a
rich DTD like DocBook that is appropriate for our software
business. From that source we can derive Postscript for
hardcopy and HTML for online delivery. I'm not saying that
everyone must operate this way, but I'll bet we aren't
alone. I'll also bet most of the information that goes
into WWW was authored in some other form and was converted
to HTML.
I believe maintaining SGML structure online is important,
but more for delivery than for authoring purposes.
The design should consider how users will want to manipulate
the information using the structural information.
At the most recent Davenport meeting, Hewlett Packard
presented their proposal for a Semantic Delivery Language
(SDL), which is an information delivery DTD. They gave up
trying to get all the HP units to author with the same
tools, and designed a common SGML format for online
delivery. By creating filters from the commonly used
authoring formats to SDL, they will be able to merge
information from a variety of sources. That is very similar
to what WWW is trying to do.
bobs