On Wed, 4 Feb 2004, Professional Software Engineering wrote:
For a home user managing their own computer, who the hell wants to have
to manage a separate host just for mail processing? Hell, for a small
business, who wants to do that?
Unless your mail volume is pretty large, there's no reason to have another
host for spamd. Horror stories of spamassassin load are usually from
people who have set something up wrong. (Refer back to "difficult to
install" in my earlier posts.) We have spamd on the same 500MHz PIII that
runs sendmail, imapd and several other services; 2500-3000 messages per
day go through spamd with no problem. (About 50% spam, I see, having
never looked at the numbers closely before.)
A mail processor shouldn't consume more resources (cpu AND memory) than
_ALL_ of the other processes on a multiuser host providing web, pop,
ftp, and other services. Yet, SA manages to do this, and some people
don't see that as wrong.
Of course it's wrong, but it's also not a true characteristic of SA when
installed and used properly (which includes not feeding it enormous
messages -- spamc won't let you do that, but plain spamassassin will).
SA might have early exit tests
It doesn't, because of the way scoring works. It was tried, but found to
produce incorrect results.
I don't know whether SA has options for skewing the results of checks
based on whether the sender is in a greylist -- such as a recognized
mailing list where you might permit a bit more lattitude WRT content
discussions, while still choosing to check for spam.
To say it "has options" would be misleading. You[*] can always plug in
new rules, e.g., to deduct points from the score for messages having
certain characteristics, which could include who the sender is.
[*] If you're an admin, that is. Refer back to "difficult to ...".
Besides, if I've got to run procmail to take action on the SA-inserted
headers, I may as well just use procmail to do the checks. Plus, if
you've got to get intimate with SA to tweak it just right, you're
already investing more time than a "canned" solution should require.
It's not *necessary* to tweak SA. As usual it's the 80/20 problem -- SA
will kill 80% or more of the spam right out of the box. This causes
people to get bent out of shape about what still gets through, so they
expend a disproportionate amount of effort to try nail that other 20%.
_______________________________________________
procmail mailing list
procmail(_at_)lists(_dot_)RWTH-Aachen(_dot_)DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/procmail