Hi,

I've got a small Postfix/Spamassassin/Procmail system which is
configured to bounces all mail that isn't for a valid local user
(default setup I believe). The "problem" is that it's also bouncing
stuff that is obviously spam but I'm puzzled as to how (or the best way)
to stop this, whilst still bouncing real emails (we have several users
who's names are often mis-spelt and so we do want genuine customers to
know if their mail hasn't been delivered).

The way I've configured my setup, Postfix doesn't do any spam testing
and instead I've wholly used Spamassassin for both RBL as well as Bayes,
et al, testing for spam by using a Procmail recipe to pipe the mail
through Spamassassin. Invalid local users are being rejected by Postfix
before getting to my Procmail recipe in any case.

I'm a bit reluctant to simply enable RBL checks in Postfix as it's a
simple pass or fail. I like Spamassassin's scoring so that periodically
anything marked as spam is (humanly) checked and any false positives are
dealt with. To make Spamassassin scoring as accurate as possible I'd
also still want to keep the RBL checks in Spamassassin, so I'm doing RBL
lookups twice for each message.

Is it therefore best practice to create a catch-all local user for
Postfix to deliver "invalid user mail" to and then modify the Procmail
recipe so that after piping through Spamassassin I bin anything
identified as spam for this catch-all address and send a preformatted
"bounce" email to anything identified as ham?

I tried Googling for a suggested solution but only came across enabling
RBL checks in Postfix.

Anyone any thoughts?

Thanks, David

David wrote:
> Hi,
>
> I've got a small Postfix/Spamassassin/Procmail system which is
> configured to bounces all mail that isn't for a valid local user
> (default setup I believe). The "problem" is that it's also bouncing
> stuff that is obviously spam but I'm puzzled as to how (or the best way)
> to stop this, whilst still bouncing real emails (we have several users
> who's names are often mis-spelt and so we do want genuine customers to
> know if their mail hasn't been delivered).
>
> The way I've configured my setup, Postfix doesn't do any spam testing
> and instead I've wholly used Spamassassin for both RBL as well as Bayes,
> et al, testing for spam by using a Procmail recipe to pipe the mail
> through Spamassassin. Invalid local users are being rejected by Postfix
> before getting to my Procmail recipe in any case.
>
> I'm a bit reluctant to simply enable RBL checks in Postfix as it's a
> simple pass or fail. I like Spamassassin's scoring so that periodically
> anything marked as spam is (humanly) checked and any false positives are
> dealt with. To make Spamassassin scoring as accurate as possible I'd
> also still want to keep the RBL checks in Spamassassin, so I'm doing RBL
> lookups twice for each message.
>
> Is it therefore best practice to create a catch-all local user for
> Postfix to deliver "invalid user mail" to and then modify the Procmail
> recipe so that after piping through Spamassassin I bin anything
> identified as spam for this catch-all address and send a preformatted
> "bounce" email to anything identified as ham?
>
> I tried Googling for a suggested solution but only came across enabling
> RBL checks in Postfix.
>
> Anyone any thoughts?
>
> Thanks, David

You can stop Postfix from rejecting unknown local recipient addresses with:

local_recipient_maps =

Then, you can tell Postfix to send all email addresses to unknown recipients
to a single local UNIX account of your choosing, with the "luser_relay" configuration setting.

See: man 5 postconf

--
Greg

Greg Hackney wrote:
> David wrote:
>> Hi,
>>
>> I've got a small Postfix/Spamassassin/Procmail system which is
>> configured to bounces all mail that isn't for a valid local user
>> (default setup I believe). The "problem" is that it's also bouncing
>> stuff that is obviously spam but I'm puzzled as to how (or the best
>> way) to stop this, whilst still bouncing real emails (we have several
>> users who's names are often mis-spelt and so we do want genuine
>> customers to know if their mail hasn't been delivered).
>>
>> The way I've configured my setup, Postfix doesn't do any spam testing
>> and instead I've wholly used Spamassassin for both RBL as well as
>> Bayes, et al, testing for spam by using a Procmail recipe to pipe the
>> mail through Spamassassin. Invalid local users are being rejected by
>> Postfix before getting to my Procmail recipe in any case.
>>
>> I'm a bit reluctant to simply enable RBL checks in Postfix as it's a
>> simple pass or fail. I like Spamassassin's scoring so that
>> periodically anything marked as spam is (humanly) checked and any
>> false positives are dealt with. To make Spamassassin scoring as
>> accurate as possible I'd also still want to keep the RBL checks in
>> Spamassassin, so I'm doing RBL lookups twice for each message.
>>
>> Is it therefore best practice to create a catch-all local user for
>> Postfix to deliver "invalid user mail" to and then modify the Procmail
>> recipe so that after piping through Spamassassin I bin anything
>> identified as spam for this catch-all address and send a preformatted
>> "bounce" email to anything identified as ham?
>>
>> I tried Googling for a suggested solution but only came across
>> enabling RBL checks in Postfix.
>>
>> Anyone any thoughts?
>>
>> Thanks, David
>
> You can stop Postfix from rejecting unknown local recipient addresses with:
>
> local_recipient_maps =
>
> Then, you can tell Postfix to send all email addresses to unknown
> recipients
> to a single local UNIX account of your choosing, with the "luser_relay"
> configuration setting.
>
> See: man 5 postconf
>
> --
> Greg
>
>
>
>
>
Thanks Greg, yes I can see how to achieve this - I was really asking if
that is the way most would recommend doing it? How do you stop your
systems from bouncing spam only?

Thanks, David.

David wrote:

> How do you stop your systems from bouncing spam

For the most part my home doesn't bounce mail, it discards it.

But it keeps a copy of all mail in a file. I run a daily report against
the Postfix syslog file for discards/rejects, and if any look like false positives, I examine
their content in the saved file, and resend the mail and repair the blocking as necessary.

However, my system is somewhat unique in that most of it's email is fetched
from POP3 accounts at various ISPs, and injected into Postfix using Postfix's
XCLIENT feature. The copy is not made by Postfix itself, but is made before being injected
into Postfix.

I do run Postfix RBL checks on both the injected email and the mail arriving directly
via the Internet. I guess I've done it long enough and adjusted configurations long enough
that there are seldom any false RBL positives. If I do find one in the logfile analysis, I'll
just fix the block, and ask the sender to resend their message.

At work, where there are millions of messages, there's not enough time in the world for
administrators to manually examine a trash pile. It's just assumed that eventually the
sender will complain to their admin or counterpart at your company, and the issue will
work it's way to your desk.

--
Greg