[fetchmail-devel] BUG: RFC violations by translated mails

[Matthias A. <ma...@dt...>]

Hi,

in certain circumstances, fetchmail will send a mail to the user. This
might be, for instance, the warning that an oversized message has been
left on the server, or that fetchmail had been unable to log in (at
least in daemon mode).

The problem is that some translations, for instance the German
translation, contains national characters, which are emitted to the mail
(Subject and other) verbatim, and it is real fun because the character
set used depends on the locale, IOW, it may depend on whether I start
fetchmail from mlterm or from xterm...

This gets really interesting if SpamAssassin is in use. Now watch this,
particularly the X-Amavis-Alert and X-Spam-Status headers: (in case you
wonder, it is de_DE.UTF-8, default on German installations of SuSE Linux
9.1, I've changed org to ork to avoid spam):

| Return-Path: <fetchmail-daemon@m2a2.dyndns.ork>
| Date: Wed, 18 Aug 2004 12:58:03 +0200 (CEST)
| Subject: Fetchmail-Warnung: übergroße Nachrichten
| Message-Id: <200...@me...>
| From: fetchmail-daemon@m2a2.dyndns.ork
| To: undisclosed-recipients: ;
| X-Virus-Scanned: by amavisd-new at m2a2.dyndns.ork
| X-Amavis-Alert: BAD HEADER Non-encoded 8-bit data (char C3 hex) in message header 'Subject'
|   Subject: Fetchmail-Warnung: \303\274bergro\303\237e Nach...
|                               ^
| X-Spam-Status: Yes, hits=7.6 tagged_above=0.0 required=6.0
|  tests=MSGID_FROM_MTA_SHORT, NO_REAL_NAME, SUBJ_ILLEGAL_CHARS
| X-Spam-Level: *******
| X-Spam-Flag: YES
| 
| Die folgenden übergroßen Nachrichten verbleiben auf dem Mail-Server pop3.web.de:
| 	1 msg 87025 Oktetts lang von fetchmail ausgelassen.
| 
| --
| 				Der Fetchmail-Dämon

Evidently, a default SpamAssassin 2.6X installation considers this
oversize warning spam. Oops!

There are several solutions to attack this:

#1 request transliteration from translators.  This won't work for
   languages that don't use a Latin character set but, for instance,
   Cyrillic, Greek or Thai to name just three. It works for some
   characters, those I know:

   ä=ae, ö=oe, ü=ue, ß=ss (sz in special cases when the preceding vowel
   is long and ss would be ambiguous), æ=ae, å=aa, ø=oe, œ=oe, ç=c

#2 request that national characters be pre-encoded as RFC-2047 (for the
   Subject) or as quoted-printable for the body. Major hassle for the
   translators.

#3 let fetchmail do proper MIME encoding of Subject and body.

I'd think #3 is the best but most complex option. 

What do you think?

-- 
Matthias Andree

NOTE YOU WILL NOT RECEIVE MY MAIL IF YOU'RE USING SPF!
Encrypted mail welcome: my GnuPG key ID is 0x052E7D95 (PGP/MIME preferred)