LuKreme, > I've thought about going with amavis (mostly for spam rejection, I > don't run VirusOS computers) but my problem has been that it seems > that either amavis runs and rejects spam (in which case I need to set > the threshold high) OR spamassasin runs and tags spam (in which case I > can leave the threshold at 5.0).
amavisd can do either, depends on how you configure it. > What I want is all 5.0+ mail to be tagged and rendered safe > (report_safe 1) and then all 10+ scoring spam to be rejected at the > transaction phase. Is this possible with Amavis? $sa_tag2_level_deflt = 5; # add 'spam' headers at that level $sa_kill_level_deflt = 10; # triggers spam evasive actions Usually amavisd is placed as a post-queue filter, where it is too late to reject mail. On a small site with not a lot of mail traffic it may be possible (i.e. it happens to work, but does not meet all Postfix requirements for a transparent proxy) to use it as a pre-queue (Postfix proxy) filter, in which case it can do what you want - additionally you need: $final_spam_destiny=D_REJECT in such a setup (this setting must not be used in a post-queue setup!) > Of course, the real problem is that I can't do per-user Bayes on the > messages in the SMTP transaction phase, Indeed, per-user Bayes is currently not possible when SpamAssassin is called from amavisd. For a user population which is not very large and diverse, I'd claim that a global Bayes performs just as good, and possibly better: all users benefit from little training. > so is it also possible to check them after the fact for bayes, > or does that just mean rerunning all of SA's tests? It would require re-running SA tests. I wouldn't bother. Mark
