Message-Id: <199710280938.KAA00502@sema.fr>
From: "Alain Zahm" <zahm@sophia.sema.fr>
To: "Keith Moore" <moore@cs.utk.edu>
Cc: "Jamie Zawinski" <jwz@netscape.com>, "Graham Klyne" <GK@acm.org>,
        <drums@cs.utk.edu>, <moore@cs.utk.edu>
Subject: Re: Can different messages have the same Message-ID 
Date: Tue, 28 Oct 1997 09:48:08 +0100
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit

Your assumptions on security are right, I don't want to start discussions
on that area, means exist to protect mail.
If Unique Message Identification is out of scope of DRUMS, then how can I
build a Messaging system that takes a real
responsibility of message delivery. How can I track message relay and
identify where it eventually get lost?

Received-lines may be a mean, but if the message is lost it could be
difficult to read them... If routers logs keep trak of Messages that passed
under their responsibility, fine we could have a mean to improve our
service quality.

Now if Message-ID is not unique and then not an identifier. I would propose
then to mention that in the RFC and then specify that Message-ID is purely
informational and by thus optional.

To answer to DJ Bernstein, no heuristic could help, we need deterministic
methods.

I know that means to provide unique identifier exist, but the decision is
more a question of responsibilty: do we think that it is our responsibility
to allow message Identification, or is it out of scope?

regards
-------------------------------------------------------------------------
Alain Zahm
SEMA GROUP Telecoms
email: zahm@sophia.sema.fr
phone: +33 (0)4 93 95 40 66
fax: +33 (0)4 93 95 40 33
Les Algorithmes, le Pythagore Bat. A
Route des Lucioles
06905 Sophia-Antipolis Cedex
France
--------------------------------------------------------------------------

----------
> De : Keith Moore <moore@cs.utk.edu>
> A : Alain Zahm <zahm@sophia.sema.fr>
> Cc : Jamie Zawinski <jwz@netscape.com>; Graham Klyne <GK@acm.org>;
drums@cs.utk.edu; moore@cs.utk.edu
> Objet : Re: Can different messages have the same Message-ID 
> Date : lundi 27 octobre 1997 22:29
> 
> The answer is simple:  it is out of scope for DRUMS to define this.
> There has never been support for per-message billing in Internet mail.
> 
> If you depend on message-id for billing, someone will duplicate
> message-ids to get by without paying.  If you suppress messages with
> duplicate id's, that creates a nice hole for denial of service attacks
> -- if I know that a particular message with a particular message-id is
> being sent to you, I can send you a different message with the same id
> and perhaps prevent you from getting the original message.
> 
> Keith
> 
> 
> > When I want to propose a commercial service for email where I charge
> > customers on a per-message basis. I would like to show a detailed bill
> > where each message is clearly identified. When my service is used as a
> > relay, I want to be able to charge remote server that use my
facilities.
> > Each mail is store in order to have a complete track of the activity
and
> > find a message if any sender or recipient complain.
> > 
> > What is the best information to rely on ?
> > 	-Date? several machine have not their date and time correctly set
> > 	-Subject? many mails share the same subject.
> > 
> > I believe Message-Id would be a nice mean to distinguish between severa
> > mails. Then standard should not allow two messages to share the same
ID, in
> > any case.
> > 
> > What is an identifier that does not allow message identification?
> > 
> >