To: drums@cs.utk.edu
Subject: Re: timezones & Date 
In-Reply-To: Your message of "Sat, 06 Apr 1996 08:37:21 PST."
             <v03005b0ead8c4d25a862@[205.214.160.121]> 
Date: Sun, 07 Apr 1996 04:53:31 +1000
Message-Id: <5387.828816811@munnari.OZ.AU>
From: Robert Elz <kre@munnari.oz.au>

    Date:        Sat, 6 Apr 1996 08:37:21 -0800
    From:        Dave Crocker <dcrocker@brandenburg.com>
    Message-ID:  <v03005b0ead8c4d25a862@[205.214.160.121]>

    Second, the process of "composing" a message can be quite
    protracted, making it impossible to use this definition and have it applied
    deterministically.  Composing can start on one day and end the next month.
    We need a definition which involves a discrete transition event, such as
    starting composition or posting for delivery.

Why?

Obviously any particular implementation will want to be
deterministic, but I can think of no reason at all that
the standard needs to be in this area.

If it did, I'd want the "start of composition" time, not
"user pressed send button", as that time is the one that
really places the messages in sequence.

Eg: consider this message - it is a reply to Dave's.  Other than
a few other of his messages in other threads, I and messages that
preceded Dave's, I have seen no other messages (no replies to
Daves).  None existed when I set out to create this reply.

Anything making a sequenced message thread out of this, should
put my messages immediately after Dave's - in a group with any
others that have also been sent in response to Dave's but not
having seen other replies.

Now its possible that since I started typing this, some other
messages have appeared in my mailbox - if the Date added to this
message is (as it will be, as that's how MH works) the time I
press send, you (and your software) have no way at all to know
whether I saw any such messages before I pressed sent this message,
in fact, it would look as if I should have.

However, if the date on this message were from when I started typing
it, there were no other messages yet, and you could tell I had seen
no others - further, you could tell from the list's Received header
that's added when this (and other) messages pass through list
expansion what the time span was from compostition start to
being redistributed, any messages that fall into the same gap
would clearly have been written in parallel with each other, and
could be so marked.

However, others really like the "time send was pushed", it allows
other information to be imparted - eg: you can tell that I was
at my workstation (well, one of them) and active at that time, and
if you receive it quickly, you have a fair chance of being able to
reply (or send an unrelated message) and have me see it quickly, or
you could call and hope I will answer the phone, etc.

Thus, this should probably be a user controllable option, so people
can set whichever they prefer - but that's more than DRUMS need
bother with, all that is relevant is that it be clear that the
Date: header relates to the time the human who composed the message
interacted with it in some way, and not to something hidden behind
the scenes (like when the message finally left the author's
workstation for the world beyond).

kre

ps: initial composttion, and "pushed send" aren't the only
reasonable options, another might be the time that the user last
started composition mode - as in, if I get to here, save the
message, then come back to it again tomorrow, and add some more
text, the Date: could reasonably be the time I started adding that
text (assuming I then proceed to transmit the message).