Re: Bounces vs. new messages

On Tue, Jun 03, 2008 at 06:59:31PM -0600, Stephen Warren wrote:

> That shouldn't be true; there's nothing in SMTP that requires a bounce
> be generated solely in the context of the SMTP transaction that is
> receiving the message being bounced.

Perhaps not, but if the sending server doesn't get bounced *during* the
transaction, a bounce would need to be a new message. This is often a
problem in complex mail systems where internal relays attempt to bounce
messages, and the new bounces are then considered backscatter or relayed
messages. Qmail suffers for this in many enterprise deployments that
I've worked on, which is one reason I tend to use postfix
professionally. YMMV.

I'm willing to stipulate that there's no reason a bounce message can't
be generated AFTER receipt, making TMDA's implementation legitimate. I
think my point was that, unless there's an RFC prohibiting treating a
bounced message as a relayed message (as opposed to an original
message), this sort of ISP behavior may be idiotic but perfectly
legitimate from an RFC standpoint.

Since most large ISPs don't really care about customer satisfaction
(they're monopolies in most areas), even *with* an RFC requirement it
would be difficult to get them to comply with sane behavior. Ever see
how many large companies (Yahoo, for example) are listed on
rfc-ignorant.org?

Without an RFC requirement, there really isn't much of a grounds on
which to get them to change at all. That's why I think this workaround
is important.

So, having said that, if Charter is basing their "unauthorized relay"
decision simply on the existence of the null envelope, is that incorrect
or just annoying?

-- 
"Oh, look: rocks!"
	-- Doctor Who, "Destiny of the Daleks"