On Thu, 2004-01-15 at 11:47, Eli wrote:
> Yeah there is - you can have a spam router that uses the accept router and
> uses BSMTP to send the message to spamc/spamd. Then add a command that runs
> exim to re-inject the message back into itself for final delivery. Your
> router would have a condition that would check to make sure it had never
> scanned the message (check a header) as well as you can specify a new
> transport name when you re-inject it that you can test on as well (people
> chose spam-scanned it seems).
>
> There is an example here:
> http://dman13.dyndns.org/~dman/config_docs/exim-spamassassin/node12.html
That's the setup I initially reviewed and found to feel to hackish, for
exactly the reasons mentioned in your summary. Maybe I'll have another
look at it. Is anyone else on this list using this method?
> I initially used this before using exiscan. It works very well, except you
> need queue runners to do final delivery, and of course you will be
> unconditionally accepting all spam, which makes it hard to do bouncebacks
> (lots of talk about this everywhere - which is why lots of people choose
> exiscan or sa-exim to do scanning before accepting the message).
I use other things at SMTP time to reject messages, but because I only
block based on the strictest of tests, I find that SpamAssassin still
tags a good number of messages. I don't plan on boucing anything after
the fact. Just delivering into a seperate directory. But the users
each need to be able to control their own scoring.
Since 99.9% of my users don't have white/black lists, or custom scoring
for individual tests. They just want to be able to set the score where
SA considers a message to be spam (or set it so high it never consider
anything spam). I'm considering using exiscan to run SA to just add
headers. Then in the local_delivery test the score header against a
file in the user's home directory containing the score at which they
consider a message to be junk, and then putting it in the right maildir.
--
Chris
