Marcin Krol wrote:
>> and would block only a tiny %age of spam.
>
> This may be so, but I would like to see some realistic numbers and
> experiments to show that. My own (admittedly, quick) estimates based on
> my mail logs indicate that the volume of spam cut short by callouts is
> approx. 5/6 of spam, while SpamAssassin has to deal with 1/6 remaining.
Without knowing for sure what mails are spam and what are ham,
we reject connections:
10% rdns + 2mx
35% helo
5% dnsbl
20% conn-dropped during delays
10% no such user
3% sender-verify
Cheers,
Jeremy
