huge bug

[Ken Murchison]

Ken Murchison wrote:
> OK.  After talking with Rob and with Mark Crispin (author of UW IMAP), 
> the problem comes down to an unfortunate implementation decision in 
> Eudora and a change in the wording of the MIME specification from RFC 
> 1521 to RFC 2046.
> 
> Technically, there is nothing syntactically incorrect with the MIME 
> message generated by Eudora.  Unfortunately, Eudora 5.1 and earlier 
> generates MIME boundary strings which can contain other boundaries as 
> substrings.
> 
> There is wording in RFC 2046 (which wasn't present in RFC 1521) which 
> states that such boundaries should be treated as the same.  This is what 
> Cyrus does.
> 
> UW IMAP whose original MIME parsing code predates RFC 2046, still 
> follows the RFC 1521 text which mentions nothing about the substring case.
> 
> None of us know why the substring text was added in RFC 2046, but both 
> Cyrus and UW can claim that they are correct.  The fact that Eudora has 
> been changed in v5.2 to match RFC 2046 is a good thing, and all users 
> are urged to upgrade.
> 
> In lieu of this, I have just added a "eudora_mime_hack" option to Cyrus, 
> which will be available in v2.1.14 ("real soon now") to work around the 
> MIME boundary issue.

OK.  Final update before we put this to rest.  Last night we discussed 
this issue with Ned Freed (author of MIME), and although we all agreed 
that Eudora is broken, we decided to be more liberal in what Cyrus allows.

So, the fix for Eudora will now enabled by default in 2.1.14 (the 
eudora_mime_hack option is gone).  We have added a rfc2046_strict option 
(default=off) to re-enable the old behavior, if necessary.

-- 
Kenneth Murchison     Oceana Matrix Ltd.
Software Engineer     21 Princeton Place
716-662-8973 x26      Orchard Park, NY 14127
--PGP Public Key--    http://www.oceana.com/~ken/ksm.pgp