Errors-to: owner-drums@CS.UTK.EDU
Date: Thu, 14 Sep 1995 20:07:46 -0700 (PDT)
From: Mark Crispin <MRC@CAC.Washington.EDU>
Sender: Mark Crispin <mrc@Tomobiki-Cho.CAC.Washington.EDU>
Subject: Re: summing up: the meaning of the reply-to header 
To: Robert Elz <kre@munnari.oz.au>
Cc: drums@CS.UTK.EDU
In-Reply-To: <17592.811133459@munnari.OZ.AU>
Message-Id: <MailManager.811134466.5293.mrc@Tomobiki-Cho.CAC.Washington.EDU>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; CHARSET=US-ASCII

On Fri, 15 Sep 1995 12:50:59 +1000, Robert Elz wrote:
> You're looking at this backwards, what is relevant here is
> not what the recipient does with the message, what is relevant,
> and all that can be relevant, is to indicate what the sender
> means when she puts in the Reply-To header.
>
> The problem currently is that Reply-To is ambiguous, it might
> mean almost anything, no-one knows, which makes it close to
> useless (ok, perhaps that is a little overboard) - the idea is
> to make it clear to the sender just what they are saying when
> they insert that header.

<Begin anecdote>

Abraham Lincoln was once asked to endorse a law that, while he agreed with its
intent, he felt was unenforcable.  To the assembled group in front of him, he
asked: "If we call the tail of a sheep a leg, then how many legs does a sheep
have?"

"Five," answered the group.

"No," Lincoln replied sadly, "because just because you call a tail a leg
doesn't make it a leg."

<End anecdote>

RFC-822 says (4.4.3.) that this field indicates the mailbox(es) to which a
reply should be sent, and specifically endorses three uses:
	1) an author who lacks a regular machine-based mailbox and wishes to
	   indicate an alternate machine address
	2) the author indicating additional persons to receive replies
	3) mailing list insertion of a Reply-To to force replies to the list.

It isn't clear what precisely (1) means.  I believe, from my memory of those
days, that the idea was if the author was on a PC that didn't have an SMTP
server.  Back in those days, I guess that we thought that the PC's name would
appear in the From: and/or Sender: lines.

We may consider (3) to be abuse of Reply-To, but it is clearly sanctioned by
RFC-822 and you'll find list managers crying "foul" if this practice is
banned.

An all-too-common use of Reply-To that has come us has been by people who have
found themselves unable to alter the From header of their messages due to
local administrative cretinism.  I can not tell you how many individuals have
*insisted* to me that this is a "necessary security feature".  It is a long
and tiresome issue, and I personally am unwilling to fight it.  Losers will
lose, no matter how much you try to win.

And there's the problem.  The poor schmucks who have to use systems managed by
losers find themselves caught if they need to redirect replies to their mail.
Every time some system imposes this policy there is a chorus of demands from
users for a capability to insert a constant Reply-To header in all outgoing
mail.  And so the technological war continues.

Maybe this is an abuse of Reply-To, but if you deny this to those users, they
too will cry "foul".

Or, more likely, both sets of (ab)users will ignore any specification that
purports to ban their practice.

I fear here that the horse is out of the barn, and that attempting to close
the barn door now is an exercise in futility.

Thus my recommendation is to punt.  Document, perhaps, but punt on trying to
fix it.