We went live on iRedMail yesterday, all has gone exceptionally well and I thoroughly appreciate what a great package iRedMail is.
But spam is a problem... for my users the only difference they can see between our old mailserver and our new mailserver is that suddenly we're getting more spam in our inboxes!
Difficult to know how much spam the greylisting is consuming, from my past experience I would say - alot - but then we used grey listing before as well. But, looking at the remaining spam, less than 10% is being correctly identified. This is undoubtedly because of the lack of bayesian analysis.
I have found this[1] page, but it does warn that it is not yet ready for production. I am trying to decide whether to wait, to ignore the warning and charge blindly on, or just re-implement what I've had working on our old server - just switching on the bayesian analysis, and running manual learning.
Do we have an ETA on this being completed? Any suggestions or advice?
Thanks!
]]>