Robin Bowes wrote:
I use bogofilter and I don't get *any* false positives!
Bully for you then! ;-)
:)
bogofilter is another one of those solutions which was designed to work well for individuals, but doesn't scale terrifically well for large numbers of people (for one thing because it relies on BDB). It also requires an MDA (like procmail) for correct operation (whereas I can run dspam inline with the regular vpopmaildeliver).
The scaling issue ran through my head as I pressed send...
The other thing is that your accuracy is largely determined by the kinds of e-mail _you_ get. Even the bogofilter docs discuss false positives (and why you might want to continuously train). If you have never gotten a false positive, then you are decidedly the exception, rather than the rule.
I generally set conservative defaults and continuously train. Spam goes to a spam folder. Undetected spam is dropped in an undetected spam folder and processed by a script running from a cronjob. Misdetected spam is processed similarly - not that I've got any :) Actually, I did used to check my Spam folder regularly but never found anything so stopped bothering!
So, you're saying I'm exceptional? :)
R. -- http://robinbowes.com
