Matt, 

You're being too nice.  It set my habeas rule to -0.01.  You never know
when you only .01 away from making it a spam :)

Gary

-----Original Message-----
From: Matt Kettler [mailto:[EMAIL PROTECTED] 
Sent: Tuesday, March 16, 2004 5:19 PM
To: John Andersen; [EMAIL PROTECTED]
Subject: Re: Habeas (again)

At 03:59 PM 3/16/04 -0900, John Andersen wrote:
>Virtually all the pill pushers getting thru in my set up are useing
>X-Habeas-SWE headers now, and getting a -6 score as a result.
>
>I'm going to have to kill this rule.
>
>I report them, but they are in brazil and probably out of Habeas's
reach.

None of this is even remotely new. They've been doing this for at least
two 
months now.

Quite frankly, I'm absolutely shocked to hear that there's anyone on the

list who hasn't reduced the score of this rule, or zeroed it out yet.
For 
all of those that haven't yet, I'd strongly suggest doing so until SA 3 
comes out, or the spammers abate their abuse (unlikely).

I've been running with the rule set to -0.1 for quite a while.. Just
enough 
to be of some minor benefit to ham, since my autolearn threshold is 
slightly negative, but not enough to be of any benefit for spammers.

If you use bayes I'd also suggest adding the habeas headers to 
bayes_ignore_header statements. This will prevent bayes from considering
it 
as a nonspam token. Prior to the outbreak, only nonspam had it, so a lot
of 
pill spams were getting low bayes scores on top of the habeas benefit.



Reply via email to