The new feature in plesk to train spamassassin worked a while for me, trained a lot of messages but now i comes with an error when i want to train. So i decided to train spamassassin server-wide with ham and spam myself at the commandline. 1) Saved some ham and spam from my outlook at my hdd 2) Upload it to the server 3) Then train like this It says: Learned tokens from 100 message(s) (100 message(s) examined) Looks good! But now i wonder if i'm training spamassassin server-wide with this method? Does this really work? Anybody have some experience with this?