Hi,
We use SA with MailScanner and delete high scoring spam. This system
works great, but is quite a resource hog with a lot of mail.
In an effort to optimize things a bit, I'm wondering if there any way to
tell Spam Assassin that once a certain score is reached, it should stop
it's checks and return?
Something like - if score > x stop testing this message and return the
current score, rather than continuing to loop through all the tests and
rbl checks. I confess I haven't looked into the code to see if this or
any permutation of this is possible. Perhaps just skip the rbl tests if
the messages scores > X ?
Also, caching of SA scores for messages with the same checksum that come
through the system within X minutes would be a big help. This would help
with dictionary attacks that have the SAME checksum. I know this is
easily worked around by spammers, but it's still very common, otherwise
systems like DCC and Pyzor wouldn't be in use. Also, some systems,
including ours, use sendmail's queue group functionality to split
recipients so that per user rules work in MailScanner/SA - this results
in identical copies of the message being submitted to SA very frequently.
Any thoughts?
Thanks,
Ken A
Pacific.Net
We use SA with MailScanner and delete high scoring spam. This system
works great, but is quite a resource hog with a lot of mail.
In an effort to optimize things a bit, I'm wondering if there any way to
tell Spam Assassin that once a certain score is reached, it should stop
it's checks and return?
Something like - if score > x stop testing this message and return the
current score, rather than continuing to loop through all the tests and
rbl checks. I confess I haven't looked into the code to see if this or
any permutation of this is possible. Perhaps just skip the rbl tests if
the messages scores > X ?
Also, caching of SA scores for messages with the same checksum that come
through the system within X minutes would be a big help. This would help
with dictionary attacks that have the SAME checksum. I know this is
easily worked around by spammers, but it's still very common, otherwise
systems like DCC and Pyzor wouldn't be in use. Also, some systems,
including ours, use sendmail's queue group functionality to split
recipients so that per user rules work in MailScanner/SA - this results
in identical copies of the message being submitted to SA very frequently.
Any thoughts?
Thanks,
Ken A
Pacific.Net