[Mimedefang] Re: calling action_bounce() for viruses

[Sevo Stille]

David F. Skoll wrote:

> That can be very difficult.  For example, our company (Roaring Penguin)
> uses a completely different mail server for outbound mail than for
> inbound mail.  The two servers are even on different ISP's.  It would
> take a fair bit of effort for our inbound server to "know" that a bounce
> message should go to someone else.

A 5xx reject by the receiving MTA will be dealt with by the outbound. In 
the standard scenario of an access provider, the users deliver to the 
outbound MTA, which should know the ASMTP id of the sender, and could 
direct its bounce messages the right way.

Mail service providers that permit relaying from in- to outbound for 
their users would have to pool ASMTP ids among their mail servers, 
and/or add a header with some authentication id to permit the outbound 
to address the true sender.

> Furthermore, some third-party systems actually generate bounce
> messages (rather than SMTP failure codes) depending on how their
> network is set up.

Technically, this is only inevitable where the mail is bounced more than 
one system away - that is, when the local recipient is valid but an 
alias forwarding to some external system that rejects the mail. If 
"undeliverable" mails are only generated in that particular case, most 
of the issue would be resolved, as in spam and virus bounces it is the 
local recipient address which is invalid.

True, there are sloppily set up multi-stage systems which cannot check 
for a valid recipient at the perimeter, but as things are developing, 
these systems are growing rogue. With fake sender spam bounces taking up 
more bandwith than our useful traffic for several days earlier this 
month, misdirected bounce messages are now a real bandwidth and cost issue.

Sevo

-- 
Sevo Stille
sevo at radiox.de