ietf-smtp
[Top] [All Lists]

Re: SMTP Transferred-By-Reference

2007-11-13 08:01:25

On Tue, Nov 13, 2007 at 07:12:38AM -0500, David F. Skoll wrote:

Alex van den Bogaerdt wrote:

The big difference seems to be what happens after you decide a
message to be spam.

Present day: you delete it without notification or you send it "back",
generating backscatter.

Or you fail it with a 5xx SMTP failure code.

A: the message submitter (a zombie)
B: A's ISP
C: you/your ISP
D: me

Scenario 1a:
A sends to B, B accepts.  B tries sending to C, C rejects.
B sends a NDN to B.

Scenario 1b:
A sends to B, B accepts.  B tries sending to C, C rejects.
B forwards the spam to /dev/null

Scenario 2a:
A sends to B, B accepts.  B tries sending to C, C accepts.
C sends a NDN to B.

Scenario 2b:
A sends to B, B accepts.  B tries sending to C, C accepts.
C forwards the spam to /dev/null


Things get more complicated when more relays are involved.

With TBR:

A sends to B, B accepts.  B informs C, C does nothing.
B kills the message eventually and does not send a NDN to D.

Things stay the same when more relays are involved.



With TBR: you send a notification to the sender's domain. That server
should NOT forward the bounce to an innocent victim. Instead, the domain
could even count such bounces and be on guard about this customer.

Rewriting the envelope sender to detect legitimate bounces vs. backscatter
can be done without an ESMTP extension.

Yeah, except that you seem to be looking at this from D's perspective.
I'm not. The key factor here is that the return path has to be at the
originator's ISP server. Or so I think.