Re: SpamAssassin - how about bogofilter?
On Wed, Oct 16, 2002 at 09:42:20AM +0300, Tarmo Toikkanen wrote:
> I was just wondering, if anyone's tried a statistical approach (e.g.
> bogofilter) to filter spam?
As I understand it, Bayesian filters rely on people reporting both spam
and nonspam to the filter. That works well for a single user, but I
think it'd work very poorly for the BTS. A lot of BTS mails never even
get a response never mind careful spam/nonspam tagging, so the quality
of the filter would suffer as a result.
> One major advantage of bogofilter (in addition to being more accurate
> than spamassassin) is that it's blazingly fast.
We aren't really hurting master with spamassassin at all - it took half
an hour to process a couple of weeks' mail, IIRC.
Also, the logistics of getting user reports through to bogofilter are
more difficult to manage than just letting SA do it. :)
> So I was just wondering if this has been discussed or tried out?
Adam did consider it. I know I preferred the spamassassin approach, but
I don't know if that's why he chose this.
--
Colin Watson [cjwatson@flatline.org.uk]
Reply to: