Re: [listmaster] Lowering SA score for your list
> On 04.10.07 15:55, Martin Zobel-Helas wrote:
>> some listmasters are currently monitoring the amount of spam to your
>> list, and see quite a big amount of spam coming to your list.
>>
>> Since beginning of September we have the possibility to set SA scores
>> per list. I hereby would like to ask, if it makes sence for you, that we
>> lower the SA score from now 3.6 pts to 3.4 points and see what happens.
>>
>> Be aware that lowering the score by .2 points might risk in having more
>> false possitives, which then would not appear on the lists.
>>
>> Please report back to listmaster@lists.d.o if you think that this is
>> reasonable. Please reach consensus on the list first.
>
> I personally think that spam fight should be done on the global level, I'm
> subscribed to many lists so a bit less mail oin one of them would not help
> me.
>
Me too. Hooking the MTA to spamhaus and other blacklists would certainly
help.
On the local level, I have found using bayes tokens to be very efficient
and actually allows me to increase SA scores, lowering the probability of
false positives. When my bayesian database is robust enough I even set
BAYES_99 to 12.0, giving spam messages scores above rejection. This lowers
the CPU usage considerably.
Ricardo
Reply to: