Re: [exim] Reducing Spam Assassin Load

Peter Bowyer wrote:

>On 02/10/05, Steve Lamb <grey@???> wrote:
>
>
>>Marc Perkel wrote:
>>
>>
>>>What I want to do is after a few that Exim learns somehow that this is
>>>all ham and just bless it without having to rerun SA every time. Exim is
>>>at least 1000 times faster than SA and bypassing SA reduces system load.
>>>
>>>
>> I understand what you wanted, Marc. I was not replying to that at all. I
>>was just pointing out that Peter was incorrect in his equation that
>>SpamAssassin is Bayesian and Bayesian is SpamAssassin. Nothing more than
>>that, nothing less.
>>
>>
>
>Please show me where I outlined such an equation? Come on, I don't
>mind being told I'm wrong when I am, but this is going a bit far.....
>I did refer solely to the Bayesian capabilities of SA without
>mentioning that it does other things as well, but believe me, I'm very
>much aware of that.
>
>My point (possibly poorly expressed) was all about watching out that
>you don't burn just as many cycles working out how to bypass SA as you
>would simply letting it do its job, and at the same time, throwing the
>Bayesian baby out with the bathwater. We've been round this loop a few
>times ourselves.
>
>Peter
>
>
>
Peter - where you are wrong is that SA has VERY high overhead. It's a
perl program running thousand of regular expressions, network lookups,
bayesian comparison, and learning. I love SA but avoiding it is a huge
increase in performance. Especially if the avoinding it is simple.

People use things like greylisting and such to avoid using SA to reduce
load processing spam. But no one has done anything to reduce load
processing ham. Every good message runs through SA and I think there's a
way to bless some of that especially where the same message is coming in
for lots of users. This would really cut don't peak loads.

I like the idea of accessing the AWL of SA possibly to see if I can find
ways to bypass SA based on reputation. What would be ideal is to do some
fort of hints database with a short term reputation that expires in hours.