Greylisting if not contentious, at the very least is not a one-shoe-
fits-all-sizes nicety. It's not a good choice for some/all/many.
Greylisting also breaks mail that is automated. In many models if
there's not a human willing to resend cr@p back to permit their
connection the email is SOL. As such it's a practice ripe with failures.
Greylisting relies on "The assumption is that since temporary failures
are built into the RFC specifications for e-mail delivery, a
legitimate server will attempt to connect again later on to deliver
the e-mail." Unfortunately that's not a valid assumption as mail need
not be queued in the SMTP model. Hence it's a very dangerous practice.
And the RFC actually doesn't require redelivery attempts so it's a
very erroneous assumption.
But sure, it will reduce the amount of mail you get. A lot of it
legitimate.
A better way of rejecting spam is robust caller sendout verification.
Unfortunately postfix's implementation, while it will work and reduce
spam, is much weaker than other MTAs such as exim, which also utilizes
very robust ACLs for all phases of the processing.
-dhan
------------------------------------------------------------------------
Dan Shoop
Computer Scientist
iWiring / U.S. Technical Services
