Sql based Spamassassin with sieve

[Ted Cabeen]

Ramprasad A Padmanabhan <ramprasad at netcore.co.in> writes:

> On Thu, 2003-09-11 at 05:53, Ted Cabeen wrote:
>> Ramprasad A Padmanabhan <ramprasad at netcore.co.in> writes:
>> 
>> > I am using a gateway spamassassin with Mailscanner. The problem is every
>> > user would like his own individual preferences stored.
>> >
>> > So I was looking if there was a way I could use sieve to fork
>> > spamassassin per user with the user name and the prefernces stored in
>> > Mysql database.
>> >
>> > I found that some people have done this, But there doesnt seem to be
>> > much documentation
>> 
>> We do it with amavisd-new before the mail reaches cyrus.  It uses a
>> slightly different schema for the preferences than SA, but it works
>> well.  Using amavisd-new for this also means that multi-recipient
>> messages are only run through SA once, regardless of the number of
>> recipients.
> Precisely.
> But thats what I do not want. There are some people in my office who
> themselves will decide what is spam and what is not
> Now How Can I run a spamassassasin with individual settings
> If there are 5 rcpts the SA must sperate instances run for all 5 of them
> Must be someway in sieve But I am not able to get one working

Why does it need to run 5 times?  If you've got 5 people with
different settings, you can run spamassassin one time and still have
individual preferences.  The score of a message is not determined by a
user's SA preferences in most cases.  In general, every recipient will
get the same score for the message.  The only thing that changes is
whether the message's score is sufficient to tag the message as spam.
amavisd-new is designed around this method.  It runs the spam check,
gets a result and then compares the score of the message to the
threshold for each recipient.  Depending on their threshold, the
message is either tagged as spam or not.

Do many of your users have SA preference requirements that they would
get a different score for a message than other users?

Also, for whitelists and the like, you can just add sieve rules for
the whitelisting before you check the X-Spam-Flag header.

-- 
Ted Cabeen
Sr. Systems/Network Administrator
Impulse Internet Services