Re: Processing after the end of DATA

> > Rejecting anything that you can at the SMTP level, rather than 
> > generating a bounce, strikes me as a sufficiently Good Thing that it 
> > outweighs any remaining benefits from the "MUST minimize" rule.  
> > Minimizing bounces seems more valuable than minimizing SMTP 
> > connection time, primarily because the bounces are more likely to 
> > waste an actual human being's time at some later point.  -- Nathaniel
> >   
> Agree with this, but I still don't run my spam filters before the 
> final ACCEPT.  As long as there is a possibility that the filter will 
> take too long, it isn't worth the confusion resulting from a timeout.
Almost all of the commercial enterprise-grade filters I know do *all* of 
their filtering before accept: spam, virus, content policies, etc. 
That's because of the desire to emit a clear rejection, and backscatter 
is a far more serious problem than duplicates. The exceptions (those 
that do filtering after the 250 OK) don't generate DSNs; they deliver to 
a spam folder or other quarantine.
<csg>