[RFCI-Discuss] (no subject)

[Alan Brown]

On Tue, 18 Oct 2005, Matus UHLAR - fantomas wrote:

> > Not at all. It comes down to "Don't accept it in the first place without
> > checking deliverability"
>
> There still are (and imho there always will be) cases when you accept mail
> and later find it undeliverable. It may look (and even be) deliverable at
> the time you check validity of RCPT TO:

In the case of "quota exceeded", simply holding the DATA transaction
open while the LMTP process runs will prevent most bounces.


As for Qmail, it's been a menace since it was released. Apart from its
built-in accept-then-think-about-it policy, it sends one recipient per
SMTP transaction (*) and it opens up as many SMTP sessions as there are
messages to deliver to a given host (**)

(*)  This drove up costs to one satellite-phone connected customer
     in the Cook Islands by 5000% (yes, 50 times the cost) and is
     spamming in its own right (bandwidth waste, comparable to Usenet
     EMP spam)

(**) It is a denial of service attack when qmails may try to open 100+
     SMTP transactions simultaneously. I know that later ones could be
     limited to "only" a few, but it's still abusive.