Use MailScanner and Spamassassin spamd

Matt Kettler mkettler at
Wed Nov 22 01:02:08 GMT 2006

Res wrote:
> On Tue, 21 Nov 2006, Matt Kettler wrote:
>> Leah Kubik wrote:
>>> I am wondering if it might be possible to run spamd on a server that is
>>> running MailScanner and not have major issues?
> Yep.
>> Why? it would be considerably slower.
>> MailScanner uses SA at the API level (ie: Mail::SpamAssassin), and
>> does so on a
>> persistent basis, therefore it acts as it's own spamd. Calling an
>> external
>> process would be considerably slower.
> I think the issue is so undividual users have preferences, rather than
> putting up with system wide tuff luck settings.

True, but you won't get MailScanner to use that spamd. However, you can run
MailScanner at the MTA layer without SA enabled, and call SA at the MTA layer
(ie: via procmail). At that point MS would only handle virus scanning.

You'd gain the advantage of per-user configs, but also loose the performance
gain of one-scan per message. Now you'll be doing one SA scan per recipient,
which shouldn't be that big a difference, but it does add load when a spam gets
bcc'ed to 10 people at once. It's all part of the standard trade-offs of MTA vs
MDA layer scanning. (And yes, there are tricks to make MS handle recipients one
at a time, but you still don't get per-user configs)

That said, as it turns out none of this has anything to do with what the OP
wanted. They wanted to remotely update a bayes DB using spamd as a target of
bayes learning. (A new feature of spamc in 3.1x is the ability to to "spamc -L
spam <message.txt")

More information about the MailScanner mailing list