Upgrade spamassassin

Matt Kettler mkettler at EVI-INC.COM
Fri Jul 11 21:06:26 IST 2003


At 01:27 PM 7/11/2003 -0500, Joe Stuart wrote:
>thanks, it worked. The weird thing is that I have old spam that I'm
>sending to myself

There's a few reasons that sending yourself old spam may cause a negative hit.

1) by sending the message to yourself, you've changed the message headers.
This removes a large quantity of information that SA uses to look for
"invalid" patterns commonly present in spam, such as date fields that
contain a non-existent timezone, and replaces them with legitimate headers,
which can greatly reduce their score.

2) in general I've found the 2.5x ruleset to be slightly weaker than past
versions, but this weakness is greatly made up for by the awesome power of
the bayes engine. Once you get bayes trained 2.5x works considerably
better, and more to the point, it's harder for spammers to tune their
emails to avoid it (because everyone has a different training).

3) Also how old is "old"? In general the current version of SA is tuned to
catch current trends in spam. Since the behavior of spammers (and
legitimate mailers) changes over time, the current version of SA might not
catch all historical spam. But that's ok, because you really want SA to
catch the spam you get today. How well it catches spam you got a year ago
isn't very important.



More information about the MailScanner mailing list