From: "Eric S. Raymond" <esr@snark.thyrsus.com>
Message-Id: <199710031949.PAA04283@snark.thyrsus.com>
Subject: Re: wishful thinking vs. reality
To: mau@beatles.cselt.it
Date: Fri, 3 Oct 1997 15:49:47 -0400 (EDT)
Cc: drums@cs.utk.edu
In-Reply-To: <199710031449.QAA15744@beatles.cselt.it> from "Maurizio Codogno" at Oct 3, 97 04:49:22 pm
Organization: Eric Conspiracy Secret Labs
Content-Type: text

Maurizio:
> But then I considered that we are not expliciting what *people* must do,
> but what *UA* must do. Therefore it makes perfectly sense to dictate
> that the UA MUST put a specific address in the To: field when replying,
> even if the user could change it at her will, and even if nobody
> can be sure if the UA is really compliant.

Well put!

I was going to make the same point.  Would everybody please try to
bear this in mind in further discussion?

The new Reply-To language would be improved by language specifying
that it is a requirement on the *UA*, not the user.  I therefore
modify the Raymond/Zawinski proposal (as endorsed by Dave Crocker) to
read:

------------------------------- CUT HERE --------------------------------
Mail agents typically support one or more "reply" commands which automatically
assemble reply address lists from the headers of the message being replied to.
Most support both "personal" replies (to the From address only) and "group"
replies (including the contents of other headers such as To, Cc, and Bcc).

       o  If the "Reply-To" field exists, then all reply lists 
          assembled by the mail user agent MUST use the contents of the
          Reply-to field in place of the contents of the From field.

       o  The contents of Reply-To SHOULD NOT replace any other
          fields (such as To, Cc, and Bcc) in assembling reply lists.

       o  User agents may (and typically do) permit users to edit address
          fields in a reply; the presence of Reply-To MUST NOT prevent
          or limit such editing, nor prevent nor limit any other form of
          explicit user-specified redirection.

------------------------------- CUT HERE --------------------------------

The philosophy of this proposal (which, for further discussion, let us
label "Raymond-2") is that we stop trying to sweep the fact that we are
putting requirements on the UA under the rug.  After all, 822 already did
this -- otherwise the distinction between To and Cc wouldn't make sense.

Dave, is the addition of the third bullet point acceptable to you as
a correct gloss on the original 822 language?

Partisans of other positions: So we son't wrangle forever, please try to
develop counterproposals in this format.  Explain how your proposal differs
from this one.  If your proposal differs in substance from the 822 language
(so that it might be considered new practice) justify that.

I hope that explicitly bringing the UA design into the discussion, and
specifying that the Reply-To field controls the UA's default behavior
but *not* the user's choices, will permit us to resolve this thorny
question.
-- 
		<a href="http://www.ccil.org/~esr">Eric S. Raymond</a>