I like Keith's proposal a lot, particularly in its choice of
start-new-stuff and end-new-stuff characters, which aren't
meta-characters at all. I had thought that Erik's proposal was
excellent except for possible prior use of his ``*<n>'' codes, and in
fact Erik's could be superior in that it provides a graceful method by
which folks may give their name in Romanized fashion as well as in
full-blown arbitrary-charset form.
Is it worth making the Romanized spelling of a name one of the
parameters of a Keith-form:
=?1?Q?J&o/rn?Jorn?=
(more compelling in non-Roman languages, to be sure).
Does Erik's proposal apply to Subject: lines as well? E.g.
Header-Encoding:
1=text/iso2022=base64=gd0p4Jq0N2Yt09jU524c1p239049023849023842390849032
Subject: Apologies *1
Craig