http://issues.apache.org/SpamAssassin/show_bug.cgi?id=5179





------- Additional Comments From [EMAIL PROTECTED]  2006-11-12 15:36 -------
FWIW:

This may or may not pertain, as the subject of "horizontal" whitespace versus
"vertical" whitespace has been mentioned... I don't want to confuse the issue
regarding the specific header/body separator issue presented here, but the
DomainKey and DKIM standards have had issues in the past with MTA munge
(sendmail), where trailing whitespace in the header portion of an email is
modified, causing the signatures to fail.

http://sourceforge.net/tracker/index.php?func=detail&aid=1485150&group_id=110311&atid=656974

It may be noteworthy that my testing has been specific to DKIM running in
relaxed/simple mode (relaxed for header canonicalization due to sendmail munge,
and simple for body canonicalization). I've chosen this method (relaxed/simple
versus simple/simple) over using _FFR_ANTICIPATE_SENDMAIL_MUNGE since it's my
understanding that relaxed canonicalization is "production-ready" code, instead
of being for future release.

If anyone feels it'd be valuable for me to put my signing systems in
simple/simple with _FFR_ANTICIPATE_SENDMAIL_MUNGE to see what effect it has on
the header/body separator issue illustrated here, I'd be happy to do so. It may
be of limited value, though, as the reflector hosted at
[EMAIL PROTECTED] does DKIM in simple/simple mode and it verifies
fine with my hack to milter-spamc (and doesn't without it). The DomainKey from
the same reflector doesn't verify for me, but only because it wasn't signed with
h= style DomainKey headers and because I add a header to the email before it
hits SpamAssassin - this is my fault. The DKIM standard uses h= style signatures
by default, so header additions in the chain don't break the signatures.

Also noteworthy is that DK_VERIFIED otherwise works great... I am using nofws
mode to sign my messages with DomainKey and sendmail + milter-spamc + SA +
Mail::DomainKeys has never had any trouble on the verification end. This could
be due, however, to the idea that the header/body separator is not part of the
DomainKey canonicalization.

As far as MIME goes, it's never been an issue. I purposely test signatures with
HTML emails to "stress-test" the system into parsing all that extra body data...
and it's never been a problem. I'm guessing because the MIME data goes beyond
the header/body separator it's a non-issue (at least for my purposes).



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to