To: drums@cs.utk.edu
Subject: Re: wishful thinking vs. reality 
In-Reply-To: Your message of "05 Oct 1997 00:04:26 GMT."
             <19971005000426.22898.qmail@cr.yp.to> 
Date: Sun, 05 Oct 1997 10:57:45 +1000
Message-Id: <19367.876013065@munnari.OZ.AU>
From: Robert Elz <kre@munnari.OZ.AU>

    Date:        5 Oct 1997 00:04:26 -0000
    From:        "D. J. Bernstein" <djb@cr.yp.to>
    Message-ID:  <19971005000426.22898.qmail@cr.yp.to>

  | Occasionally RFC 1123 goes beyond the protocols into user-interface
  | design, where automation is certainly relevant:

For 1123 one can argue that that is OK - 1123 is hosts requirements
(or half of it), and is attempting to specify what is a conforming
internet host.   One might argue (I'm not saying that I would) that
a host without a mail UA offering a "reply" function is not useful
enough to the internet to be considered a conformant host, so in
such a doc one may want to specify some UI functionality.

(1123 also includes protocol updates, which makes it a bit of a mess,
but the updates were needed, no-one would want to attempt to conform
to a protocol agreed to have defects).

However for 822bis this is irrelevant.   822bis is not "requirements
for internet mail user agents", it is the end to end protocol spec for
user to user e-mail.

If someone wants to go write a requirements for internet mail user
agents, then go do that - but don't attempt to make a mess out of
822bis in order to make 822bis be that document.

And to Dan - no, I was not disagreeing with anything you said, so there
is no need to reply and berate me.

But...

  | Fortunately, this sort of paternalism is now prohibited by RFC 2119.

while agreeing that we don't want it, I would point out that 2119
doesn't prohibit anything.  No-one is required to use the definitions
in 2119 for anything at all, docs can define their own terms, or simply
not use any like that if they choose.   Using 2119 is nice, where
applicable, because (assuming it is used properly) keeps all the various
specs reasonably consistent, but if its definitions don't suit the
purposes of some new doc, 2119 can be ignored.

kre