On Wed, 2009-08-05 at 02:07 +0000, alxgomz@??? wrote:
> Very good link thank you...
> a) and b) solution seems really unacceptable to me... c) solution is better.
> But i think i'll keep using spam checks in transport and routers because
> even if the c) solution exposed in the link *CAN* handle per_user policy
> based on SA "required_score" It sounds like this score cannot be calculated
> using per_users test such as autowhitelist or bayesians.
c) doesn't really do per user settings, it just uses the site global
settings - ie site AWL and site bayes db - and then using a
system_filter or .forward filters to decide what to do based on how many
stars are present in the X-Spam: Yes (+++++) header.
a) and b) both allow for an specific SA profile to be used that includes
AWL and bayes DB only for that user/profile.
If you want to do single user scanning, you need to have a message to
scan that is only headed to that person which is achieved in exim by
deferring all recipients that don't fit the same settings.
Once a message has made it to the DATA ACL, all the recipients have been
accepted and there is no way to signal the sending MTA that we want to
reject some of the recipients leading to either an accept and bounce,
quarantine, or mark and pass on system which are less than optimal.
--
The Exim manual -
http://docs.exim.org
