Re: [Asrg] New proposal for spam blocking: Greylisting

[Vernon Schryver <vjs@calcite.rhyolite.com>]

> From: Evan Harris <eharris@puremagic.com>

> ...
> > That makes sense for a small site, but I wonder if AOL or Microsoft
> > would agree.
>
> The fact that they have systems that can handle those volumes of mail means
> that they also probably already have the needed infrastructure in the way of
> distributed scalable database backends.

It's one thing to have machinery that can send a half a billion messages
per day with currently typical queue delays.  It like something quite
different to have machinery that could send 500M messages/day, many
retransmitted 1-3 times more than currently and spending 1000 times
longer in the queue.

If all of AOL's peers used your system, a lot of AOL's mail would
start spending 10s or 100s of minutes instead of seconds in AOLs queue.

The cost of a retransmission is about the same as the cost of sending
a message.  Given an average message size of ~5K bytes, most of the
round trips as well as all of the DNS and TCP SYN delays in an SMTP
transaction are burned by the end of the Rcpt_TO command.  Contrary
to common naive calculations, the 6 packets and 3 or 4 round trips of
the DATA command are often the tail of the dog.


> In the overall picture of handling any individual mail, the checks for
> greylisting is a tiny amount.  Multiply that by a million mails an hour, and
> it's still the same small percent, which means it should be straightforward
> to implement it in a manner that scales with whatever infrastructure is
> already in place.
>
> Places like AOL and yahoo already use databases to store their user
> accounts, and all the contacts for those user accounts.  Reimplement it to
> take advantage of those existing databases, and the impact becomes even
> less.

I'm not talking about the costs for the recipients running your system
but for the senders.



> > My bet is the opposite and that the employers of legislators will
> > never let them outlaw "mainsleaze" even as they instruct them to outlaw
> > the current spammers.  To prove the point at an extreme, consider the
> > likelihood of congresscritters outlawing their own spam.
>
> So let them be handled by blacklists.  In either case, the fact that they
> are trying to be legit means they are easily blocked since the headers
> should not be forged. ...

Blacklists do not work for "mainsleaze" because a significant part
and usually most of those spews is not spam because it is wanted by
its targets.


Vernon Schryver    vjs@rhyolite.com

_______________________________________________
Asrg mailing list
Asrg@ietf.org
https://www1.ietf.org/mailman/listinfo/asrg