I resisted automated mail filtering for a couple of years, figuring that the problem wasn’t too bad, and that I could always detect and delete spam with only a little bit of work.
Then my e-mail address spread far and wide in the credits files of a few software projects and dictionary attacks became cheaper and… then I took everyone else’s good advice and installed SpamAssassin. Now I skim a folder full of questionable mail for a few seconds a day and train false negatives once a week and don’t worry about spam as a user. (As an administrator of the mail server, I worry about the waste of resources, but that’s a separate problem.)
I wish I didn’t need this software, but I do, and it works. Thank you to all of the contributors to SpamAssassin!
Oh, and for everyone curious about my spam training aliases, they are:
alias learnspam='sa-learn --spam --mbox ~/Mail/spamtrain && sa-learn --spam --mbox ~/Mail/questionable'
alias learnham='sa-learn --ham --mbox ~/Mail/hamtrain'

You're welcome! It goes both ways; we couldn't have done it without all the work you've done on perl ;)
my question is do you train it on mail messages that have been marked up using spamassassin? Is that a problem, if so?