[NTLUG:Discuss] Postfix Spam Blocking

[Victor Brilon]

Brian wrote:
>>>If I could get more accurate SA scores, that would
>>>solve most of my spam issues!
>>
>>Can you explain what you mean by that please?
> 
> 
> Sure...obvious spam, already learned as spam, continually scored at
> less than zero, which then gets autolearned as ham.  I go back,
> re-learn it as spam (according to docs, this will "unlearn" it has
> ham).  Repeat cycle.

Uh....this is most definitely not the default SA configuration. First of 
all, I literally have never seen a spam score less than 0. Second of 
all, a score of less than 0 does not mean the message is autolearned as 
ham either.

> Or, spam which continually makes it through SA is sent back through
> sa-learn.  One would think after several learnings of identical spam,
> you'd at least get a BAYES_50 hit.   

Yes it should. Have you checked your logs to see what SA says about 
learning messages?

> I'm working on upgrading to 3.0.2...so any complaints from me up to this point
> are no longer valid :)

I'd also suggest wiping out your Bayes DBs when you upgrade. I have a 
sneaking suspicion that they somehow have become corrupt.

Let us know how the upgrade goes!

Victor