"glyph" doesn't seem like the right term...but I'm not sure what is. Maybe:
"...an algorithm for converting an octet stream into characters" is
sufficient.
Another approach:
* define a "character set" as a set of (description, code) pairs, where the
"description" says "how to display a character" (without being too
specific) and the "code" says "how to represent this character in an
octet-stream".
* The code has to be unique -- you can't have more than one description
for a given code, though a description can describe multiple possible
ways to display a character like "broken vertical bar"...it's a fine line.)
* The code also has to be recognizable with no look-ahead.
(One problem with this approach is that it doesn't include code-switching
sequences, so obviously it still needs work. Charsets with combining
characters could be fit into this scheme, as long as the sequences of
combining characters were always in a well-defined order, the definition
specified by the RFC that defines the particular charset.)
I would also prefer that the MIME2 document not specifically exclude "bare
10646"...especially since we don't know what IS 10646 will be yet, and
simply because this might be confusing.
Keith