From: Craig McLean [mailto:[EMAIL PROTECTED] > > In addition to all the good responses you've had I would add that if > you keep the spam after learning, as I do (in case the bayes DB get > killed) then it can sometimes take a while for sa-learn to finish, > even though it will skip messages it has already seen. I learn from > 5 mailboxes (ham and spam) and recently hit a problem where one > sa-learn had not finished by the time the next one started, and > failed because the DB (using file-based db) was locked.
What I would do is something like this (untested): TEMPDIR=`mktemp -d` mv /spam/mailbox/dir/* $TEMPDIR sa-learn --spam $TEMPDIR mv $TEMPDIR/* /spam/archive/dir/ rmdir $TEMPDIR 1) Move the messages to a temporary directory so you don't have to worry about new stuff coming in while you're working. 2) Run sa-learn on the temporary directory. 3) Move the learned mail to an archive directory for holding This way sa-learn does not have to continually parse the list of messages that it has already learned and all of the old messages are still available for re-learning if you need to rebuild the database. Bowie