Re: Bayes isn't working

2007-06-18 Thread Martin Strand
On Mon, 18 Jun 2007 20:06:52 +0200, Rob Wright [EMAIL PROTECTED]  
wrote:



sa-learn --ham -C /etc/mail/spamassassin --showdots --spam --no-sync
Maildir/new


Don't use the --spam flag when learning ham


Re: How Do I Enable RBLs

2007-06-14 Thread Martin Strand
On Fri, 15 Jun 2007 01:51:50 +0200, Michele Neylon :: Blacknight  
[EMAIL PROTECTED] wrote:



John Rudd wrote:

LuKreme wrote:

On 10-Jun-2007, at 16:54, Peter Pluta wrote:

  reject_rbl_client zen.spamhaus.org
  reject_rbl_client list.dsbl.org,
  reject_rbl_client bl.spamcop.net,
  reject_rbl_client sbl-xbl.spamhaus.org


Er, no.  zen OR sbl-xbl.  I've found spamcop to hit far too much ham  
for my tastes, and I never found that dsbl was hitting anything (or at  
least nothing that sbl-xbl (now zen) didn't already catch).


 I do zen and dsbl, and dsbl catches about 1 for every 20 that zen  
does.  I do both _just_in_case_ there isn't perfect overlap.
  I agree entirely about spamcop.  Some people use it for spam marking,  
which I am also leery about ... but it seems to me to be absolutely  
insane to use spamcop for an actual block list.


It's fine for scoring against, but blocking is insanity!



I tested SpamCop for our info@ address at work (about 200 messages a day)  
and didn't get a single FP for six months.
I use it for blocking on our mailserver now (about 2000 accounts) and  
haven't received any complaints so far. :)


This guy's stats seem to confirm my observations:
http://stats.dnsbl.com/

Martin