I think that you should be allowed by the browser to INC whatever,
wherever at your own risk. If you create bad HTML this way, then the
browsers already deal with this. INC is intended to be a macro-like
facility for information providers that is subject to the same problems
and issues as the C preprocessor faces in resolving #include's. I would
expect INC to be used in a similar manner. You can theoretically
(although probably never in practice) #include <main.c> somewhere
inappropriate, and its not cpp's problem, rather the compiler's. In this
case, INC would be called by the parser to resolve a reference which
would be inserted into the parser's HTML data structure.
The only show-stopper non-DTD related syntactical problem I can forsee
facing browser implementors is a recursive inclusion loop. An
intelligent rendering agent might need to maintain an inclusion level
counter so that it could bail after descending down several tens of
levels into an infinite loop.
| An INC'd document could be treated as an SGML SUBDOC entity, but
| that means that it would parse just fine no matter what it contains.
Since we are just now approaching the point where very few people have
SGML compliant rendering systems on the web that can do SUBDOC properly
or SYSTEM and PUBLIC entity resolution, much less digest and render
several DTD's, the multiple DTD problem that you imply above will
probably be resolved outside the scope of this group, hopefully at the
same time that 3.0 or greater is standardized, when hopefully we will
have a more compliant SGML implementations available on the web client
side.
It would be a parsing error according to the current DTD to pull in a
complete <HTML>...</HTML> document in the middle of a <BODY> element,
just as it would be to do the same thing in vi with :r--you don't even
need to know emacs to do this.
| Maybe this is desirable; would you propose that a browser indicate
| that an INC'd doc is some sort of inclusion, or treat it as a
| transparent part of the parent doc?
I do not see such an indication as useful for my purposes right now and
can clearly see why I would want to turn that off is present. That
functionality could be implemented by either an attribute to the <INC>
tag or by a browser tweak. I could see where, under the future direction
that I mentioned in the proposal, it would be useful for setting off the
rendering of an included, tagged portion of another document.
We should also probably pause now and work through a development and
design framework for 2.1.
-marc
--