In Santa Fe, I think we decided that a charset of MNEMONIC means
mnemonic with a base set of US-ASCII and a quoting character of &. We
also established a convention for specifying otherwise, e.g.
"MNEMONIC+ISO-10646+64" would mean MNEMONIC using ISO-10646 as the base
charset and @ as the quoting character.
Upon reflection, I think that it is a mistake to provide the latter
escape valve. The only value of it is that it means that software
targeting another charset can avoid immediately switching charsets at
the start of a message. The drawback is that it constitutes a huge
violation of our basic goal of keeping the number of "charset" values
small. I think that eliminating this is a big gain for little pain.
I agree with you.
For a human reader "&" may not be the best quoting character
(personally, I would prefer "_"), but since MNEMONIC is a
transport mechanism in character set disguise in the first place,
and only recipients with old-fashioned UA programs are supposed
to run the risk of being directly exposed to the mnemonics--
and they will meet more severe inconveniencies from other new
facilities of the RFC-XXXX standard--I see no strong reason for
complicating RFC-XXXX by allowing other characters to be used as
quoting characters.
A small advantage of making richer character sets than
US-ASCII usable as base character sets for MNEMONIC is that
certain types of text will need significantly fewer bytes after
conversion to MNEMONIC. The text should be converted to
the recipient's native coded character by a modern UA in any
case, so the use of another base character set will not increase
human interoperability.
Can't we just say that, as used in mail, MNEMONIC always starts out with
a standard base character set and a standard quoting character? Is it
really that much easier to say "MNEMONIC+ISO-10646+64" than to say
"MNEMONIC" and put a charset change at the start of the message body?
Are you talking about the mechanism for switching between coded
character sets in Richtext? I see no big need for different
varieties of MNEMONIC inside Richtext either.
/Olle
--
Olle Jarnefors, Royal Institute of Technology, Stockholm
<ojarnef(_at_)admin(_dot_)kth(_dot_)se>