I believe that Einar could be most easily satisfied with something along
the lines of a UTF8HEADERS ESMTP extension, which would specify both
that 8 bit character are permitted in the header and that those
characters MUST be interpreted as UTF-8.
there are lots of problems with this idea. for instance, there's no
way for SMTP to know whether the recipient's MUA can deal with
utf-8 even in 2047 encoded form, to say nothing about unencoded form.
and it's not as simple as saying just use utf-8 in the message header -
there's also the question of which characters are legal and which aren't,
which ones are separator characters, which ones are white space, etc.
then consider that, except for the abliity to have non-ascii email
addresses, there's no functional advantage to having raw utf-8 in headers.
there might be a slight win to having a more efficient encoder, but
MUAs would still have to implement both utf-8 and 2047 for many more years.
let's get the details of non-ascii email addresses worked out before
we start spewing utf-8 in random places in the message header.