A
Aaron
Yes, I've tried it. It detects spam using different modules, there are
quite a few that you can download include ORBL tester, alphbeta soup,
etc. The inital setup comes with keywords filter and newsletter filter.
But like k9 and POPfile there's the bayeisian portion which will
eventually do most of the work.
The key thing to note is that you can arrange/priortise a chain of
filters, with rules to pass on to the next filter or stop if spam/ham is
detected.
The interesting thing is regardless of whether the bayesian filter gets
to filter the spam, all email is always available to be trained by the
bayesian filter. At first the other modules will do most of the filtering
,but eventually the bayesian filter becomes pretty effective and you
should decide to use it as your primary filter.
One difference between it and k9/popfile is that, detected spam is not
passed to your inbox. You can review all such mail, in the spaminilator
interface, and if there is a mistake in classifying good mail as spam,
you can make spaminilator pass it on to the email client.
In terms of effectiveness, if you set up the right filters in the right
order you can get pretty good results (no lag time for training unlike
bayesian filters), but like k9/popfile with enough mails, the bayesian
filter kicks in and you should get similar results to k9/popfile.
Aaron (my email is not munged!)