I have some queries on Bayesian filtering in SpamAssassin in latest ASL.
Tom wrote this a while ago:
"The next Up2Date has SpamAssassin 2.55, that has a self-learning bayesian component. It does not require user interaction, so it is easy to implement. We do not currently plan to put in a bayesian component that requires user feedback."
1) Is this currently implemented ?
2) Tom states "self-learning" how does this work with no user feedback ?
3) Would Astaro reconsider the "no user feedback" statement ? It seems to me a few extra options in the SMTP proxy content manager such as "Delete as Spam" etc would be easy to implement ?
I have implemented SpamBayes (see SourceForge) in my MS Outlook client and it works FANTASTICALLY. I'd say it catches 99% of all spam at a guess and I think something like this would be well worth the effort.
(I trained SpamBayes on around 600 spam emails and 800 or so "ham" messages).
This thread was automatically locked due to age.