[LINK] Solution to SPAM checking

[Adrian Chadd]

On Thu, Feb 01, 2007, Craig Sanders wrote:
> On Wed, Jan 31, 2007 at 09:53:52PM +1100, Adam Todd wrote:
> > I don't know why I haven't thought of this before, years before.  Maybe it 
> 
> because you're a feeble-brained cretin who has "brilliant" ideas years
> after they've been tried and discarded by everyone else.
> 
> > If the RX SMTP server were to check the path of incoming mail or even
> > connect to the "reply address" to ensure that it was legitimate and
> > valid and was in fact related to the DNS records for the originating
> > domain, then SPAM would vanish over night.
> 
> as usual, you are wrong and you are years behind. this was thought of
> many years ago (e.g. postfix has had a sender-verification option for
> years, as has sendmail and exim and others). and no, spam DIDN'T vanish
> overnight. that's because spam isn't a simple problem amenable to simple
> solutions.

Actually, I've seen spam levels drop after turning on this kind of verification
but not because of the received spam levels. It dropped because:

(a) the networks in question couldn't be used as spam reflectors by spoofing
    from addresses w/ the intended destination, and
(b) spam to unknown addresses suddenly didn't fill the relay/scanning system
    by having to process both the incoming spam and the outgoing "no, this
    person ain't here no more" message.

So it has a positive effect, if setup carefully.

> it's also completely useless on very busy mail servers because they're
> already struggling under the load of receiving and delivering actual
> mail, let alone making an extra outbound SMTP connection for every
> incoming message (even caching the result of the Sender address lookup
> doesn't help much). so, it's vaguely useful on small, lightly-loaded
> mail servers. which hardly makes a dent (if at all) on the spam volume.

It dropped the server load of the central AV/spam scanning gateway at my
last job. Which was pushing > couple million messages a day easy.
The server suddenly didn't have to handle huge amounts of bounce load.



Adrian