Well, it's not a simple minded use of "charset".
I don't know if the gimmick you suggest above is feasible. One of
the advantages of a fixed document charset is that it makes it easy
to convert from one character encoding to another in a simple-minded
way and not lose the meaning of numeric references. But the
encodings you suggest are large enough that it's possible no one
would convert them to anything else, and it would preserve the
numeric references in 10646.
Even if this is feasible, I'd favor leaving the requirement for
10646 in the HTML 2.0 spec, and resolving this kind of extension
in an internationalization document.
For what it's worth, prior discussion of creation of new chinese
characters seems to pop up in the archives in early Feb at:
http://www.acl.lanl.gov/HTML_WG/html-wg-95q1.messages/0457.html
http://www.acl.lanl.gov/HTML_WG/html-wg-95q1.messages/0472.html
http://www.acl.lanl.gov/HTML_WG/html-wg-95q1.messages/0479.html
http://www.acl.lanl.gov/HTML_WG/html-wg-95q1.messages/0480.html
http://www.acl.lanl.gov/HTML_WG/html-wg-95q1.messages/0483.html
-- Albert Lunde Albert-Lunde@nwu.edu