Since 10646 seems to be settling down, it is important that RFC-XXXX
handle it smoothly. There are several places where more should be
said.
1) There are both 2 byte and 4 byte forms of DIS 2. They must
be announced external to the data stream. Thus just saying the
codeset is ISO-10646 is unsufficient.
2) There are possibilities of the data stream being in either
big-endian or little-endian order which should also be announced
external to the data stream (although there are provisions for
the first "word" of the stream announcing it.)
3) There is another compaction method, UTF, UCS Transformation Method.
So a specific proposal is to add the following words:
Section 5 (page 10 in ps)
add a Content-Transfer-Encoding of UTF with properties
like 8bit (i.e. not usable now)
Section 7.1
replace ISO-10646 with UCS-2 and UCS-4 (assumed big-endian)
and possibly UCS-2L and UCS-4L (to force little-endian)
(on page 18 in ps) expand on 10646 a little
Section 7.3.1
replace ISO-10646 with UCS-2 and UCS-4 (assumed big-endian)
and possibly UCS-2L and UCS-4L (to force little-endian)
Perhaps we should provide an explicit syntax for "versions" of codesets
as it is absolutely clear that 10646 will grow new characters occationally.
With the coming likelyhood of 10646 (perpetual optimist :-), I see no
need for enshrining an anachonism from the past, Mneumonic. I would
be happy if all references to it were removed. [Of course it can
still be used with appropriate x-foo constructs.] I see it as just
another codeset that is not standardized by an international body and
that will take very large tables to implement.