Path: utzoo!utgpu!water!watmath!clyde!mcdchg!chinet!att!rutgers!ukma!nrl-cmf!ames!lll-tis!TWG.COM!dcrocker
From: dcrocker@TWG.COM (Dave Crocker)
Newsgroups: comp.protocols.iso.x400.gateway
Subject: Where to put handling information
Message-ID: <8807201652.AA16735@tis.llnl.gov>
Date: 17 Jul 88 23:34:00 GMT
Sender: root@tis.llnl.gov
Distribution: inet
Organization: The Internet
Lines: 32
Approved: post-x400-gateway@tis.llnl.gov

It has been interesting to watch the discussions about the specification
of handling information, such as non-delivery notification.  This was
a point of particular concern, during the development of the MCI Mail
system.  The basic conclusion was that per-addressee, vs. per-message,
clearly provided the best flexibility.  However, the burden placed on
users can get quite painful, so that a per-message approach to user-level
specification was chosen.  The theory was that contingent assignment of
different values was relatively rare.  (Keep in mind that the user interface
model was 1200 baud, simple ascii text.)  Given a graphics- and forms-oriented
interface, the choices could be different.

With respect to coding such information into RFC822 format, you can live
within the current standard and develop a downward compatible spec or
you can develop a specification that will break some other systems.

Anything within parentheses is completely save, according to the spec,
although many systems interpret such "comments" as personal names.
Anything in local-part is safe, except for the delivery system on the
host that interprets local-part.  In either case, any information put in
either place, when a message is sent, is likely to show up in a reply
message.  What happens when you reply to the reply?

Unless you get cooperation from all of the end-points and the gateways,
you will propagate handling information in ways that is likely to be
undesireable.

An alternate approach is to upgrade 822 and choose another port for
mail transfer.  This lets participating hosts talk with each other, without
requiring participation from all hosts.  You are then left almost only
with a problem of choosing specification syntax, a relatively minor task.

Dave