Mailscanner mqueuein trouble

[Chris Green]

>On Thu, 20 Jul 2006, Denis Beauchemin wrote:
>
>>>So in essence, S.A is good for stopping about 0.1% of it.
>>>nice and handy and warns the suer the otehr 99.9% of the time BUT because 
>>>its a warning  people will at least start to read it to see if its spam 
>>>or genuine.
>
>>I disagree with you.  SA, if you tune it right, will detect pretty much 
>>ALL spam.  But it needs lots of RAM.
>>
>>We warn users above 5 and delete mails above 20 and yesterday we managed 
>>to delete 59% of spam...
>>
>
>Then its not catching the sort of crap we see, as per my previous most of 
>it is low scoring so they see it anyway, if I was to run your settings here 
>it would be a pure waste of time because you are still sending 99% of it to 
>users to evaluate if itsspam or not, but each network to their own i 
>supose.
>
>But I see in your followup post you only do about 100K a day, we have 3 
>machines that do 3/4 of a million+  and 1 that does well over a million a 
>day (which is the problem one) each so they are not exactly little 
>machines.
>
Are you running sa-learn on those boxes? I'm truely surprised at the woeful 
performance you're getting. It's almost like you are feeding spam in to 
sa-learn as ham...!