To: drums@cs.utk.edu
Subject: Re: Headers and agents
In-Reply-To: <661qgSSUcsB@khms.westfalen.de>
Date: Tue, 2 Apr 1996 18:03:42 +0100
From: Tim Goodwin <tim@pipex.net>
Message-ID: <AAAOfzFhXe4ADUAh@pipe.pipex.net>

> There is absolutely no reason why that message should go to the list, and  

Suppose I want mail to <tim@pipex.net> to be delivered to two different
mailboxes.  Why should I *not* set it up as a mailing list, and continue
to use it as my originator address?  Clearly, in this case, bounces
should be sent to the mailing list as normal, with the null return path
preserved.

You might prefer to use a sendmail style "alias" for this, rather than
a full mailing list.  But some mail software, PP for instance, does not
offer sendmail style aliases.

(I regret to report that PP doesn't preserve the null return path when
sending a message to a mailing list.  In fact it always immediately
rewrites it to the local administrator address.  *Sigh.*)

> in fact it is *very* annoying to be on a list where this happens.

Well, I find spam annoying.  Can we outlaw that in 822bis? :-)

Seriously, if it's annoying, then something's misconfigured, or
you're using the wrong tool for the job.

> > It is a mistake at any protocol level to base routing decisions on the
> > source address.
> 
> It is? Why?

Because it is the purpose of destination addresses to drive routing
processes.  Because routing on source addresses vastly increases
complexity.  Because source addresses cannot be trusted.

The only widespread example of routing based on source address that I
can think of is IP filtering (and there the source address is only used
to decide whether to route or not).  I doubt you'll find anyone today to
argue that IP filtering is an appropriate or reliable technology for any
real world problem (although it is often convenient and cheap).

Tim.

P.S. In case there are any champions of IP filtering out there,
I'll gladly receive your flames to my mailbox, but this is entirely
irrelevant to DRUMS.  Thanks.