OK. Tried the -tD. There is says that my Bayes database only has 20 spams in it. (<200).
But I have had thousands of SPAMS. Apparently, I've only fed 20 of them for sa-learn. I seem to recall that there is a flag to set for Bayes to pick up on all Spam messages over a certain score. Have to go and review again.
Autolearning is enabled by default, with a threshold of 12.0 I think... however be aware that the conditions are a bit more strict than they sound. It's not just the score alone.
score recomputed with bayes disabled (no self-feedback) including change of scoreset.
total recomputed score must be over threshold
must have 3.0 points of header-rule hits, regardless of threshold
must have 3.0 points of body-rule hits, regardless of threshold
message must not fall into strong ham-category in existing bayes learning (ie: don't contradict your existing knowledge)
oportunistic attempt to lock the bayes DB for write must succeed first try (ie: don't logjam the mail queue waiting to learn a message, if it's already locked by another process, skip the learn)
In general, autolearning is pretty rare, especially if you don't have any network-tests (RBLs, etc) to help boost up the non-bayes scores.
