Recommended spam.lists and/or sendmail dnsbl settings?

[Steve Freegard]

Duncan, Brian M. wrote:
> I guess it is the case that everyone has different needs.

Yup - most definitely.

> We never reject messages at the MTA level. (Well actually messages that
> are destined to users that do not have valid MS Exchange SMTP records
> are rejected, so I guess never is not correct, but that is the only case
> we reject at the edge)

Count yourself very lucky then - I've worked with many companies in the 
  past that had similar policies. They got so much junk they were adding 
extra MailScanner servers or upgrading existing machines every 6 months 
or so to attempt to keep up with the load that this imposed on them.

> RBL's tend to be a love/hate thing.  We love them, based on the fact
> that we still deliver every failed RBL message to the users Junk Mail
> folder. (Giving them the option to "white list" in outlook RBL'ed
> sources.

Yes - but in the case of Spamhaus (which is why people like them) if you 
do some analysis you'll find that unless you've got horsepower, disk 
space (and the associated money) to burn it isn't worth delivering these 
messages.

 From the last SpamAssassin mass-check network tests run:

xbl.spamhaus.org hit on 68.7% of spam messages and 0.0033% non-spam 
messages (3 out of 90160 non-spam messages)

pbl.spamhaus.org hit on 61% of spam messages and 0.43% non-spam (390 out 
of 90160 non-spam messages)

sbl.spamhaus.org hit on 1.26% of spam message and 0.0388% non-spam (35 
out of 90160 non-spam messages)

Based on those stats - I love RBLs too as that tells me that I could 
potentially gain 70% efficiency by rejecting them before they get to 
MailScanner.

> Given that Mailscanner allows "high scoring treatment" on RBL checked
> messages, and then the capability to set the intended actions (including
> delivery) I would think the extra control over RBL behavior could allow
> even finer tuning in some environments.

I agree - everyone has different requirements and I think checking the 
Spam Lists in order and stopping at the first hit would make sense from 
an efficiency point of view.

Cheers,
Steve.