Fed up with spamd
I’m finally fed up with spamd from Spamassassin. Spamassassin is a spam filter project, used to filter spam either on a client or a server. It has rules-based filtering and bayesian filtering. Support for it is integrated into Evolution, the mail client that I use on Linux. However, spamd is dog slow.
A little history. Back in the day, before Evolution included support for spamd (indeed, before I used Evolution, and even before Evolution), I tried a number of spam filters. This was in the early days of Bayesian filters, and my personal evaluation settled on Bogofilter as my filter of choice. It is fast, and it’s accurate. I used it with a series of scripts. When I switched to Evolution, I carried those scripts over, and included them as filters. It still required me to save a message to a file, and run the learning script on it, when it was mis-classified, but it worked, and I didn’t get enough spam that the few mis-classifications mattered.
Evolution 2.0, tho, has built in support for spam filtering, including buttons to mark messages as spam or not-spam. This is very nice, and, realistically, I get way too much spam to to deal with mis-classifications by hand. I really need the buttons. The spam filtering in Evo works okay, but it gets a lower hit rate than bogofilter did, and it’s way too slow. For example, today, for reasons which aren’t important, I re-filtered my current LKML folder, which has 27,000 odd messages in it. I started this filtering over an hour ago, and it’s still going. In addition, it takes long enough, and is resource intensive enough, to impact my working on my desktop or reading mail when filtering incoming messages. This has got to stop.
So, I’m moving back to bogofilter. This will involve writing and testing a set of scripts to make bogofilter seem like spamd/spamc to Evo. Watch this space for progress, as I’ll continue blogging about it as I go. Hopefully, the result will be a nice howto on replacing spamassassin in Evo with something else.

Discussion Area - Leave a Comment