My default setup is to set the thresholds at 0.5 and 0.9 (i.e., everything below 50% probability of being spam is ham, everything above 90% is spam).

I then let ham pass to the proper mailbox, move spam to spam mailbox (and wade through it once a week to find any possible false positives). The trick is with unsure (the ones with 50-90% probability of being spam). I *copy* them to ‘unsure’ and treat them otherwise as ham. Then I delete them from where they end up (if they’re spam) and periodically go to unsure to train Bogofilter (this was spam and that was ham).

It took me about two weeks to train Bogofilter properly (during these I checked spam mailbox more often, as there were more false positives) but now I’m absolutely sold on it. Bogofilter filters out 1100 spams weekly, I see no false positives (as I’m filtering out only if there’s >90% probability that it’s spam), no false negatives, and around ten times in a week Bogofilter classifies some spams as unsure (i.e., in the 50-90% spamicity range), so I train it on them.

I never managed to make SpamAssassin (+ SpamBayes) even close to this efficiency.

As to why one uses a chain of applications or programs – each of the programs is designed to do one thing right. This is the Unix way of doing things, where you have applications, which are like tools in a toolbox, which you then chain together to get the desired effect.

I was not happy with the thunderbird/evolution way of doing things. With procmail, for example, I can set up filters and move mail to the right mailboxes, reducing the effort involved in organizing mail on a day to day basis. Spamassassin has its obvious benefits. Mutt is a commandline mail user agent (mail reader) and it very fast, uses very little memory, and has loads of keyboard shortcuts. The shortcuts, and the behavior in general is very highly configurable, making it more attractive that Thunderbird or Evolution. I have had several very specific issues with both evolution and thunderbird over the past 2 years, some of which are not going to change soon.

The best part of the exercise I am involved in is that I know exactly how mail is delivered and processed on the system – making it easier to manage and archive mail. WIth evolution and thunderbird, I never got to know what format the mailbox uses, where it is, and how to back it up. I did know how to backup after some research, but they all have their quirks, and a tendency to use their own specific methods, making migration difficult for example. Once I have my system setup, I can use thunderbird as a reader too, by just pointing it at the right directory holding all my mail. So I only see benefits in the long run for what involves a day’s work to set up.

Hope this clears up a few things.

1: VMWare, which I have installed for the sake of just a couple of applications, won’t run with the kernel that Breezy ships with.

2: The / partition on my main HDD is behaving oddly, giving me all sorts of errors. I’m currently using a spare HDD as my /, since my /home partition is behaving perfectly (it’s on the same disk as the hosed partition and I haven’t had the guts to fsck it yet to see if the problems have spread).

Once I’m sure that the disk’s okay and VMWare release an update for the latest kernels I’ll be upgrading quicker than a very quick thing.

I’m left wondering why? Is there something Mozilla Thunderbird can’t do?