I have to say, I really don't understand why people are insisting that all
charsets used in MIME be variants of ASCII.
Because MIME is expected to work with current mail transports, which
expect text to consist of sequences of octets with certain values
that correspond to the ASCII graphics or a few control characters,
and which consists of relatively short lines ending with 0D hex
or 0A hex or 0D 0A hex. (depending on the transport).
If you try to define a charset which doesn't fit this category,
and you still want the mail to work in the existing internet
mail transport, then some additional restrictions are required.
Likewise, many existing programs that interpret text/* body parts
(for current MIME charsets) also generally expect charsets which
are variants of ASCII. Given the requirement that an unknown
text/* body part should be treated as text/plain, it's not a good
idea to define a charset for text/* which would not work well
under these circumstances.
I agree that we need to clarify how to handle charsets that are
not derived from ASCII.
Keith