For the last year or so I’ve been quite happy with the results that Spam Assassin was able to give me when filtering my email. Lately I’ve been getting a little disappointed with it’s performance, I could upgrade to a new version, or I could switch to a new filter. I had previously shied away from Bayesian filters because I didn’t fully understand the math that was going on behind them. Now, fortunately, I’ve learned how the math works and they’re really damn cool.
So my choices were to stick with SpamAssassin and it’s Bayesian filters, but it’s written in Perl and I hate Perl. I’ve become quite the Perl bigot. It’s really an ugly language and I’ll be glad when the language dies. Time was a few years ago when almost everyone used perl, no almost no one does. Python and PHP have largely replaced it. Even though, I must admit there are cases when Perl is faster. But, I digress. I could have also used BogoFilter but decided not to. Instead, I selected the relatively small market CRM114 Discriminator. Partially because it has a cool name, and partially because Paul Graham had to bow to it’s goodness at the 2002 Spam Conference.
Installation was relatively straight forward thanks to the howto. With that running now I’m in the process of training it. At first it missed everything. Then I told it all that was spam and it started to mark everything as spam. Then I got a few more real emails and now it seems like is about 80% accurate after about 20 emails of training. We’ll see how this evolves over the next few days. If it works well I’ll move my other accounts over to using it. If not, I guess I might try bogofilter or some other program. Who knows.