Well, as I said, I'm pretty sure I'm not understanding what is going on,
but my understanding was that the prohibition on encoding headers did
not apply to anything but content-type and content-transport-encoding
for the body part(s).
I believe that the following is permitted over a seven bit transport.
If it is not, then "kludge" is probably too mild.
Starting from Greg's rewrite, with his lines that I have replaced marked
">" and my new lines marked "#"...
From: <ASCII form>
Subject: <ASCII form>
Content-type: message/iso-8859-2
# Content-transport-encoding: quoted-printable
# X-Comment: that can certainly appear in an outer header for
# "message", can't it? There is no nesting here.
From: <iso-8859-2-form> * Untransportable
# From: phrase-in-iso8859-2-quoted-printable
# <ascii-local-part(_at_)ascii-domain>
Subject: <iso-8859-2-form> * Untransportable
# Subject: iso8859-2-quoted-printable-form
Content-type: text/iso-8859-2
Content-Encoding: Quoted-Printable
Message body in iso-8859-2 character set.
# (in quoted-printable, of course)
This is what I've assumed we are talking about in a "no multiple
encodings" model. It follows the model of
--main header tells about coding and char set of the inner header
--inner header tells about the coding and char set of the message
and, at no point does no transport-encode something that is already
transport-encoded.
And we have a standing prohibition on anything but ASCII in the
Content-type and -encoding fields, don't we? And in the spelling of
such things as "From:", "Subject:",...?
If that isn't how we have defined it, can we go back and do it that way?
Even if that means a new content-type in the outermost headers? All of
the other solutions, including the multipart one Greg outlined, are too
horrible to contemplate for merely being able to write one's name or a
subject field in non-ASCII characters. The solution above is merely
disgusting but, for me at least, I can manage to contemplate it.
Note that, over 8bit transport, the arrangement above survives without
alteration, only the encodings to quoted-printable go away.
--john