[H-GEN] Re: several messages

[Bruce Campbell]

On Mon, 6 Mar 2006, Greg Black wrote:

> On 2006-03-05, Bruce Campbell wrote:
>
>> In the Humbug setup, there is a slight twist, in that email sent to the
>> list addresses will be immediately accepted if it is from a recognised
>> subscriber to that list.  Note that this test is performed on the SMTP
>> MAIL FROM field, and not on the not-yet-received 'From:' address in the
>> mail data.
>
> This test is either unimplemented at present or is implemented
> incorrectly.

Neither I'm afraid; a classic case of a working test not being used due to 
an earlier test issuing the defer message.  One slight reworking of the 
test order later, and the intended behaviour works (I've removed your 
entries from the greylisting db so you can catch me out again ;) )

>  I don't use greylisting because of the collateral
> damage of this type.

This is where we need to balance speed of delivery against cost of 
receiving the email.  Humbug is not a business and, imho, does not have a 
pressing need for immediate delivery, but Humbug does have to pay for its 
expenses, such as traffic received.

With client-side filtering on junk mail, this means that the Humbug mail 
servers do have to pay to receive traffic which Humbug members later throw 
away.  As an illustration, this is the amount of spam (in K) that my 
Humbug aliases received recently:

          80694  mail/spam-to-be-processed/2005/12
         112288  mail/spam-to-be-processed/2006/01
 	 99544  mail/spam-to-be-processed/2006/02
 	 16682  mail/spam-to-be-processed/2006/03

That figure gets repeated for most Humbug members, which even at 1c/meg, 
really starts to add up over time.

With server-side filtering before the bulk of the message is received, the 
amount of traffic received is markedly reduced, and thus, so is the bill 
that the shadowy cabal presents to Humbug every month or so.

> Yes, I understand the desire to reduce spam levels, since that
> is a constant battle that I have to wage as well.  But I will
> always whine when what I consider ill-conceived attempts to do
> that result in my email being delayed or rejected.  This is one
> of those whines.

The only time email will be rejected by the greylisting process is when 
the sending MTA gets the meaning of 'temporary' confused with 'permanent'.

Delays are far more common, and is an unavoidable side effect of 
Greylisting.  However, the delay is only on the first instance; following
mails from matching tuples are passed through much quicker.

On Mon, 6 Mar 2006, Ted Percival wrote:

> >     The initial block period is five minutes.  The tuple will be
> >     remembered for 7 days, and if they haven't sent any additional
> >     mail during that time, it will be forgotten and they will have to
> >     pass the test again.

> Isn't the tuple lifetime usually set closer to a year?

Good $DEITY no.  You'll find that once you have a lot of entries in there, 
spammers start to repeat themselves.  If you have a matching tuple, likely 
created as a result of their previous attempt to send you spam, you will 
all of a sudden receive their new round of spam.

--==--
Bruce.