From: cos@leftbank.com (Ofer Inbar)
Message-Id: <9604100342.AA26217@leftbank.com>
Subject: Re: timezones & Date
To: drums@cs.utk.edu
Date: Tue, 9 Apr 1996 23:43:20 -0400 (EDT)
In-Reply-To: <v03005f06ad90c3582a68@[205.214.160.104]> from "Dave Crocker" at Apr 9, 96 06:50:02 pm
Organization: The Left Bank Operation - http://www.leftbank.com/
Reply-To: cos@leftbank.com
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

Dave Crocker <dcrocker@brandenburg.com> wrote:
> 	I sent out an informal query to a substantial-sized list of people
> I know who are interested in email, to see what they thought should be the
> definition of the Date field.  These people are by no means representative
> of "average users".  Some are developers and some are administrators and
> some are useless consultants, like me.
> 
> 	Nonetheless, the responses were interesting.  For one thing, they
> were totally consistent.  Everyone chose option 3 (or included it as
> acceptable.)
> 
> 	To save you the expense of reading below unless you want to, I will
> tell you that options three is "when the user hit the send command".  This
> is not the same as creation, nor last-modification (though it usually is)
> nor of the postmark (though it often is) but indicates the event signaling
> the user's decision to send the darn thing.

That is my choice too.  In fact, I was just now planning to post to
drums describing why I think that is the best choice, and also to
mention in passing that I'd asked some friends who deal with Internet
mail a lot about it, and they favored this choice too.

So, my explanation, in brief:

For Date: to be meaningful, it should have a standard definition.  As
I've stated before, I think "time message was composed" is just too
vague to be meaningful.  If that's what people want, we might as well
just query the sending user, "what date would you like this message
stamped with?".  Or they could put a date at the top of the message.

Date:'s meaning is useful if a) users can, and generally do, expect it
to mean what it does, and b) software that makes use of it does so
based on that same meaning.  For the most part, both of these already
hold true for "time user caused message to be sent."  Because most
email software behaves like this, or a very close approximation of
this, today, this is what most users of Internet mail expect, today.

Because most users have this expectation, mail software that
interprets the Date: header is also designed with this expectation, if
any, in mind.  So, any software today that assumes Date: has a clear
meaning, as far as I know, assumes this meaning.

As we know, this isn't always correct.  There is some software out
there, for example, that generates Date: when the user begins
composing a message, and other software that leaves it to the first
MTA to add the Date:.  But the user expectation is natural, and it's
not likely to change.  What should change is the software.  And the
first step towards that, is making a clear statement in the next
Internet Mail RFC, that Date: means what most people think it means.

Just now, as I was typing this message, I asked two other people who I
happened to be talking to, about this.  Both of them are users of
email but are not system administrators or programmers.  I didn't tell
them why I was asking, I simply asked, "what do you think the Date:
header in an email message means?".  Both of them answered that they
thought it meant, when the user sent the message (note, not "composed"
or "wrote", but "sent").  I then asked them to clarify whether they
meant "when the user hit the send key" or "when the message entered
transport".  Both said they'd never thought about that distinction
before, but now that they think about it, it's the time the user hit
the send key.  One of the people even emphasized: "The time the user
hit the send key - it's when THEY sent the message."

I questioned both of these people independently, neither of them knew
of my conversation with the other.  I expect I'll get the same answer
if I ask another user.  In fact, let me try that...  Yup.

So, Date: already has a meaning.  We just need to state it clearly.

  --  Cos (Ofer Inbar)         --  cos@leftbank.com cos@cs.brandeis.edu
  --  The Left Bank Operation  --  lbo@leftbank.com http://www.leftbank.com
  "Some might argue that designing a network which in theory survives
   nuclear warheads but not lawsuits is engineering folly..."
      -- Sean Doran <smd@cesium.clock.org>, com-priv, 18Sep95