Message-Id: <3.0.32.19990207103404.0069d6e0@pop.dial.pipex.com>
Date: Sun, 07 Feb 1999 10:36:38 +0000
To: "D. J. Bernstein" <djb@cr.yp.to>
From: Graham Klyne <GK@dial.pipex.com>
Subject: Re: Syntax extensions in RFC 821
Cc: drums@cs.utk.edu
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"

At 21:41 05/02/99 -0000, D. J. Bernstein wrote:
>Barry Leiba writes:
>> implementations that met the letter, but not the intent, of RFC821
>
>Jon Postel said that
>
>   * it was inexcusable to _generate_ syntax extensions but
>   * it was nice to _accept_ syntax extensions.

He also said, in RFC821 (Appendix E, p48):

           2yz   Positive Completion reply

               The requested
action has been successfully completed.  A
               new request may
be initiated.

I submit that sending arbitrary data when the server is expecting a new
request does not constitute a "syntax extension" so much as a "syntax
violation".

Also, in section 4.4, the state diagram clearly indicates that data may be
sent only in response to a 3yz response.  Thus, it seems clear to me that
sending data following receipt of a 2yz status code is clearly in violation
of RFC821.


Since you have provoked me into responding, I'll also pick up on an earlier
assertion you made:

>There's an installed base of MTAs delivering millions of messages a day
>through SMTP clients that treat 2xx just like 3xx.
>
>Declaring those MTAs non-compliant, and demanding that they be changed,
>means punishing the users.

This is a non sequitur.  Declaring the MTAs non-compliant imposes no
penalty on any user for whom their existing product delivers performance
that meets their needs.  If anyone is penalized, it is the developer who,
for marketing reasons, wishes to *claim* that their implementation is
compliant.

#g


------------
Graham Klyne
(GK@ACM.ORG)