[Top] [All Lists]

Re: Do the must 'bounce' rules need to be relaxed for virus infected messages?

2004-03-26 06:01:36

From: Arnt Gulbrandsen <arnt(_at_)gulbrandsen(_dot_)priv(_dot_)no>
In-Reply-To: <7382FCA44E27D411BD4A00508BD68F950AE30F04>

Daryl Odnert writes:
But why should the SMTP protocol preclude server implementations from 
performing local policy analysis after the message has been accepted 
from the client?

Because if you want to give an error message to the party who actually 
sent you the email, you have to talk to that SMTP client. (Even that's 
not a sufficient condition, merely a necessary one.)

That is not even necessary.

If you want to give an error message to the party who actually sent you the 
e-mail you need to know the real 'reverse path'. This may have nothing to 
do with the SMTP client the message was receive from. 

The validity of the return path can often only be determined AFTER the 
message has been accepted. Once the message is accepted from the client, 
analysis can show that the message did NOT come from where it claims to 
come from. 

In such a case, it is WRONG to 'bounce' to that address. That address had 
nothing to do with sending the message. That address should NOT be 
'spammed' with 'bounces' of messages that it did not send. Doing so 
victimizes the owner of the forged address.

There are half a dozen or so places in RFC2821 that need to be changed so 
as to 'legalize' the practice that has become vital to the continued 
existance of e-mail. The practice of 'dumping' know viruses should be 
encouraged. The practice of 'bouncing' those messages should be 


please pardon my infinite ignorance, the set-of-things-I-do-not-know is an 
infinite set.