ANNOUNCE and base64 encoding woes

[Yavor Doganov]

I'm working on the ANNOUNCE functionality right now, trying to
understand and fix what appears to be broken.

Here's the problem:  We need to encode the subject of the announcement
email in either base64 or quoted printable, as per the RFC.  So, to
convert the subject "Ну, заяц", I do:

$ echo "Ну, заяц" | base64
0J3Rgywg0LfQsNGP0YYK

But, here comes the problem.  When I send a test mail with the
command

$ echo '<URL:http://www.gnu.org/foo>' \
    | mail -s '=?UTF-8?B?0J3Rgywg0LfQsNGP0YYK?=' address@hidden

Mutt displays the subject as "Ну заяц?" (with a question mark at the
end), and Wanderlust displays it right, but there's an empty line
between the Subject and the other headers and they are rendered
improperly.  When I tried to send a test message with WL (the subject
being constructed by the Emacs SEMI library), I get

=?UTF-8?B?0J3Rgywg0LfQsNGP0YY=?=

Note the missing `K' at the end has been replaced with '='.  Of
course, there is no problem parsing this header (Mutt, WL, Evolution),
and I get what I expect.

Trying to encode with perl/python/recode gives the same results as
the base64 program.  Any clues what's wrong?  I'm dying to know...

Alternatively, we may encode to quoted-printable (which appears to
work in all cases), but that will add a dependency on qprint or
recode, which are not universally available as `base64'.  Or, we
could use some Perl module to do the job, which is also less desirable
than `base64'.