From: "Bart Schaefer" <schaefer@candle.brasslantern.com>
Message-Id: <970422164609.ZM14021@candle.brasslantern.com>
Date: Tue, 22 Apr 1997 16:46:09 -0700
In-Reply-To: <199704222237.SAA08455@ig.cs.utk.edu>
Comments: In reply to Keith Moore <moore@cs.utk.edu>
        "Re: How Reply SHOULD Work" (Apr 22,  6:37pm)
References: <199704222237.SAA08455@ig.cs.utk.edu>
Reply-To: schaefer@nbn.com
To: Keith Moore <moore@cs.utk.edu>
Subject: Re: How Reply SHOULD Work
Cc: drums@cs.utk.edu
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii

On Apr 22,  6:37pm, Keith Moore wrote:
} Subject: Re: How Reply SHOULD Work
}
} It's important for a message author to be able to send a 
} message to several recipients and request that replies be sent
} ONLY to a particular set of addresses.  If Reply-to only replaces 
} those addresses in the From field, this functionality is lost.
} 
} [...] But there's no way we can 
} define the protocol in such a way that the user agent always
} (or even usually) automatically chooses the correct recipients
} for a reply. [...]

As I've said before, my concern is that if 822bis says nothing beyond
that Reply-To has the more restrictive interpretation, then user agent
implementors will find themselves under pressure (real or imaginary)
to make that behavior automatic and unalterable.  ("Why did FrobMail
permit you to reply to me?  I specifically put Reply-To: grobnitz in
my headers, and RFCxxxx says that means ONLY to grobnitz.  FrobMail is
broken, and we won't use it here any longer.")

I think that at the very least some explicit mention of this:

} The right choice inherently rests with the person composing the reply.

also needs to be included, in whatever form.

-- 
Bart Schaefer                             Brass Lantern Enterprises
http://www.well.com/user/barts            http://www.nbn.com/people/lantern