From: "Dotan Cohen" <dotancohen@xxxxxxxxx>
On 26/12/06, Tim <ignored_mailbox@xxxxxxxxxxxx> wrote:
On Tue, 2006-12-26 at 09:08 +0000, James Wilkinson wrote:
> Also, I'd strongly recommend training SA's Bayesian analysis, using
> the sa-learn program. SpamAssassin won't use Bayesian analysis until
> it has learnt 200 good ("ham") e-mails and 200 spams.
Isn't that supposed to be the point of the junk/not-junk buttons on mail
clients?
> Bayesian analysis continues to be a *very* good way of analysing
> e-mail, in my experience.
Back when I was still using Windows, I used to use the in-built one that
came with The Bat! mail client. It seemed to do a reasonable job, and
it was damn quick (unlike the speed of any kind of mail filtering in
Evolution). Though, before that, I'd knocked most spam off, without any
false positives, with about 12 mail rules.
Kmail has a very decent bayesian filter called bogomail. I had to
teach it for about two weeks before it started doing any filtering
itself, with about 70 spams a day to 20 or so hams. At first it
started filtering out only the most obvious spam, but now (about two
months after first install) it only lets 2-3 spams through a day, with
no false positives so far.
But Bogofilter is my second line of defence. On the POP3 server I've
got spammassasin running, it traps about 300 spams a day, with two or
three false positives that I've discovered in the past few months.
So my 400:20 spam:ham ratio gets down to 70:20 at the server, and
further reduced to 3:20 at the Inbox. I can live with that.
Bicycle with one pedal, indeed.
500 to 1000 emails a day to my accounts.
150 to 300 of them are spams.
Typical false markups are 0 to 3 a week.
(I run a fairly agressive set of SARE rule sets in addition to the DNS tests
and default SpamAssassin rules.)
{^_^}
--
fedora-list mailing list
fedora-list@xxxxxxxxxx
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
