ohta-san,
i've cc'ed this to the ietf-822 list, but perhaps any more
exchanges should just be between us; i can't imagine the email
folks being too interested in this..
first of all, plan 9 uses 16 bits now because that's all that
10646 defines right now. if it changes to need more bits, we'll
change our code; it's as simple as that.
secondly, although i am sympathetic to issues of bandwidth,
i'm not terribly concerned if a new japanese plane takes 4 bytes
to encode each (17 bit) character. it is only interchange, after all.
there is no compulsion to use 10646 internally within your systems,
just pressure to use it BETWEEN systems. plus, where appropriate
you can use the lower codepoints to save space (presumably the
new plane contains many duplicates from the CJK characters).
thirdly, adopting standards causes everyone pain (the 822
folks know this!). compromises are inevitable; perhaps we should
live with them for a while and take it from there..
andrew