On 11 Aug 2020, at 10:48, jin&hitman&Barracuda wrote:
Hello,
I'ld like to ask some questions about filtering/scanning mails. I'm
using
Postfix-3.4.14 in my setup. However, the questions aren't directly
related
with Postfix but, as a newbie I don't have such an opportunity to ask
this
kind of questions to anyone except this list, so please accept my
apologies.
I need a design which allows me to scan/check e-mails (the DATA
portion)
against some words stored in the SQL table. Actually, there are two
tables stored in the same SQL DB which are named "white" and "black".
As
you probably guess, white db table contains the words acceptable and
black
contains the contrary.
To what end?
If the goal is to filter "spam" (specifically: unsolicited bulk email)
then I agree with Dr. Venema: try SpamAssassin, either with the
single-purpose spamass-milter or via some more comprehensive milter like
MIMEDefang or Amavis. However you hook up SpamAssassin to Postfix, it
will not make a simple binary decision on filtering based on trivial SQL
tables as described, because that mechanism is completely inadequate for
spam filtering. SA does include a Naive Bayesian statistical filter that
can run out of a SQL RDBMS, and I suppose one could train that filter
artificially to behave much like simple allow/deny lists of words, but I
would not recommend that approach. SA is not designed or intended to
operate as a generic filter, it is only really useful for spam
filtering. (Full disclosure: I don't speak for the SpamAssassin Project,
but I am a member of the PMC.)
If the goal is the process, i.e. to use simple allow/deny lists of words
in a SQL DB to categorize messages, then SpamAssassin in any normal sort
of configuration is not the right tool. If you can write a relatively
small amount of Perl, then MIMEDefang might be. It is a general
filtering framework which CAN use SA and other filtering tools, but it
is customized by writing Perl subroutines and designed to do complex
message manipulation, so it could be used as a platform for the sort of
filtering that you propose.
When I try to find a method, I've found milter
options. There are many milters out there and they offer different
features. Looks like i have to write one to serve my requirements.
But, I
still don't know the answers of the questions listed below.
- Are there any other options that allow me to scan a mail against
values
stored in an SQL database ?
I am aware of no such tool for use with Postfix or any other MTA. For
the purposes that most people want to filter mail, that mechanism is
just too simplistic.
- If the only way is using a milter to achieve this goal, when exactly
i
should send it to the milter ? Does it matter ?
I'm not clear on what you mean by that...
A milter registers callback routines with the MTA based on what aspects
of the SMTP transaction it wants to filter. Since you are filtering the
message body, you'd need to register xxfi_body() and xxfi_eom()
callbacks if you build a custom milter. Postfix makes those calls
synchronously at the end of the SMTP DATA command, the only time when it
has message data.
- (Sorry for this one but, let me ask this specific question once.)
Instead
of writing a new milter, can I use "proxsmtp" project for that purpose
?
I'm not familiar with it, but it appears to be a SMTP proxy, not a
milter. It may or may not work with Postfix's proxy interface. Whether
you can make it filter mail in the manner you desire, I do not kn ow.
--
Bill Cole
b...@scconsult.com or billc...@apache.org
(AKA @grumpybozo and many *@billmail.scconsult.com addresses)
Not For Hire (currently)
