Re: Evolution junk filter not working
On Sat, 2005-02-26 at 22:20 +0200, Neilen Marais wrote:
> Hi Bram
>
> On Sat, 2005-02-26 at 17:43 +0100, Bram Mertens wrote:
>
> > On Sat, 2005-02-26 at 17:45 +0200, Neilen Marais wrote:
> > AFAIK Evolution only uses Spamassassin so make sure you have
> > Spamassassin installed and configured correctly. (You need a.o. to
> > edit /etc/default/spamassassin: change "ENABLED=0" to "ENABLED=1")
> >
> > And test it from the CLI (with spamassassin and spamc).
>
> Ah, thanks, it seems I don't have spamassassin installed. Do I really
> need to use spamd, or can evolution just call spamassassin?
I'll have to point out that I'm by no means an evo expert but using
spamassassin in stead of spamd really hurts performance-wise so I expect
evo will onyl use spamd.
> Anyway, I reckon Evolution should produce a warning if spamassasin can't
> be invoked. Silently failing might give people the idea that its spam
> filtering just doesn't work too well, as it did me!
Evo (or SA) writes status messages to /var/log/mail.log so If you don't
see anything like:
Feb 27 07:50:55 localhost spamd[29170]: connection from
localhost.localdomain [127.0.0.1] at port 39049
Feb 27 07:50:55 localhost spamd[29170]: info: setuid to m8ram succeeded
Feb 27 07:50:55 localhost spamd[29170]: checking message
<244702D5.044A32E@mindspring.com> for m8ram:1000.
Feb 27 07:50:58 localhost spamd[29170]: identified spam (5.6/5.0) for
m8ram:1000 in 3.3 seconds, 3894 byte
s.
Feb 27 07:50:58 localhost spamd[29170]: result: Y 5 -
BAYES_60,DRUGS_PAIN,HTML_MESSAGE,HTML_TEXT_AFTER_HT
ML,MIME_BASE64_BLANKS,MIME_BASE64_TEXT,MIME_HTML_ONLY,SARE_URI_PILLS
scantime=3.3,size=3894,mid=<244702D5.
there it might suggest a problem with your setup. But I agree, although
I seem to remember that the evo-developers decided that when SA wasn't
installed on a machine it would indicate that the user didn't want to
use it. So reporting it as an error would not be "A Good Thing"...
Perhaps you can check the evo-manual, if it doesn't mention this you
might want to report it as a bug in evo's bugzilla.
[...]
> My previous experience with bogofilter, also a bayesian filter, showed
> me that it's quite effective soon after I have started training it. I
> also found that whenever I marked a specific message as junk, it would
> be immediately recognised as such. I guess I was expecting it to work
> the same.
I don't know bogofilter but perhaps their algorithms are different?
> I assume one trains spamassin for spam by "marking as junk", and
> everything else get's trained as ham by default. Or do I have to do
> something else?
I'm not sure about training the rest as ham, I have auto-learn enabled
but mark as junk does run sa-learn on the selected messages. (BTW I've
learned that it's best to select multiple messages and mark them as junk
at once rather than one message at a time.)
[...]
Regards
Bram
--
# Mertens Bram "M8ram" <bram-mertens@linux.be> Linux User #349737 #
# debian testing kernel 2.6.8-1-686 i686 512MB RAM #
# 15:39:08 up 19 days, 19:25, 11 users, load average: 0.43, 0.36, 0.24 #
Reply to: