Sender: jwz@jwz.org
Message-ID: <36D19AFF.2C7CD682@mozilla.org>
Date: Mon, 22 Feb 1999 09:59:27 -0800
From: Jamie Zawinski <jwz@mozilla.org>
Organization: mozilla dot org
MIME-Version: 1.0
To: Pete Resnick <presnick@qualcomm.com>
CC: drums@cs.utk.edu, lc-drums-msg-fmt@apps.ietf.org
Subject: Re: References and In-Reply-To (section 3.6.4)
References: <9276.919402389@munnari.OZ.AU>
	 <v04104702b2f1fc45e496@resnick2.qualcomm.com>
	 <36CC5877.6A74F226@mozilla.org> <9276.919402389@munnari.OZ.AU>
	 <7BDhXar1w-B@khms.westfalen.de>
	 <v0410470db2f4a27b6dd1@resnick2.qualcomm.com> <v04104805b2f63cdd6ea5@resnick2.qualcomm.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Pete Resnick wrote:
> 
> On 2/20/99 at 8:07 PM +0000, D. J. Bernstein wrote:
> 
> > People also reply to web pages. Do you allow URLs in References?
> 
> Nope, because it violates 822 syntax. Multiple message id's in
> In-Reply-To does not.

And allowing multiple parents in References violates 1036 semantics.
So, we need to pick one, and in this case, I choose 1036 over 822.

> So what do we do about In-Reply-To? It certainly allows multiple
> message id's in 822 and the semantics is clear enough; I assume noone
> wants me to change In-Reply-To *again* to remove a legal 822 syntax,
> does anyone? And if I leave In-Reply-To with multiple message id's,
> what do I say about its relationship (if any) to References? And what
> do I say about multiple parents wrt References? Deny the existence of
> such a construct?

I find it hard to see changing the semantics of In-Reply-To as being a
big deal, given that in my surveys of common usage of In-Reply-To, I've
not once come across a real-world usage of In-Reply-To that contained
multiple message IDs.  I've come across In-Reply-To fields that
contained "<X@Y>" text -- but in the examples I've grepped for, it has
always been the case that only one was an ID, the rest were addresses.

Perhaps some MUAs do this, but the simple fact is, it's so uncommon that
it hardly counts as an installed base we need to be compatible with,
especially when compared to the installed base that uses 1036 semantics,
and especially when considering that this is a semantic change and not a
syntactic change, and isn't going to cause anything to malfunction.

The 1036 semantics are proven to be useful; the 822 semantics are vague
and have no proven utility.

-- 
Jamie Zawinski
jwz@mozilla.org    http://www.mozilla.org/   (work)
jwz@jwz.org        http://www.jwz.org/       (play)