Errors-to: owner-drums@CS.UTK.EDU
Message-Id: <199509131614.MAA17199@wilma.cs.utk.edu>
From: Keith Moore <moore@CS.UTK.EDU>
To: Olle Jarnefors <ojarnef@admin.kth.se>
cc: drums@CS.UTK.EDU, Keith Moore <moore@CS.UTK.EDU>
Subject: Re: Another header-munging example..... 
In-reply-to: Your message of "Wed, 13 Sep 1995 17:11:56 +0200."
             <9509131511.AA04174@mercutio.admin.kth.se> 
Date: Wed, 13 Sep 1995 12:14:15 -0400
Sender: moore@CS.UTK.EDU

> OK, drums@CS.UTK.EDU was the wrong list for a constructive 
> contribution to solving the mailing list problems. Which 
> other IETF-related list would you recommend for that 
> discussion? mailext? ietf-822?

I suggest that the discussion be delayed until more of the
drums stuff is settled.  A lot (not all) of the problems 
with mailing lists are caused by a lack of uniformity
in user agents as to how they do things like replies.
Lists try to "fix" this by munging headers, after which
user agent writers try to "fix" this by adding more reply
commands or doing reply commands differently (hence the
idea that "reply-to" only replaces the "from" header 
addresses when generating replies).

Anyway, we need to see how the generic 822 stuff settles
out with respect to reply-to and resent-* before we start
defining new header for lists. 

> > Discussion of how to add headers to 822 to fix mailing lists 
> > is unlikely to be productive.
> 
> Since you seem quite sure about this, I must ask you about your
> reasons: Is it because defining new headers can't solve the
> problems technically, not even after a transition period, or
> because the interested people will never be able to reach rough
> consensus on a solution?

It's because at present there are two camps -- the "we need more
headers" people and the "we need to fix the ones we've got"
people, and they're butting heads rather than compromising.

Personally, I see a real danger that we'll damage interoperability 
if we define more ways to do essentially the same thing (say, tell 
a user agent where to send replies), and some of the proposals for
new headers do exactly that.  We need to encourage greater uniformity 
in user agent behavior, and the best way to do that is to encourage 
a more narrow interpretation of the existing 822 headers.  

I also see a need for a new header or two to support lists.   But 
how those new headers should be defined has a lot to do with how
we decide to interpret the existing headers.

> > We should confine our discussion to:
> > 
> > + clarifying what the existing 822 headers mean
> > + clarifying the existing 822 language about how "text message
> >   teleconferencing" (see section 4.4.3) servers can use existing 822 
> >   fields like Resent-* and Reply-to
> > + clarifying how user agents should interpret 822 fields
> >   (to the extent that we define user agent behavior at all)
> 
> I don't believe that such fine-tuning can solve any of the
> important mailing list problems. The reason is simple:
> 
> The headers defined in RFC 822 were designed primarily to cater
> for personal e-mail. The use of mailing lists on a large scale
> was at its infancy in 1982. When a message is distributed by
> a mailing list exploder the recipient's UA needs more
> information about the message _in addition to_ all information
> that can be included in a person-to-person message. Especially
> the UA needs to know that the message came from a mailing list
> and the identity of that list. (Both needs can be satisfied
> by a single new header, Mailing-List.)

I agree with most of this.  The question in my mind is not whether
it's worthwile to solve this problem (it is), but whether we should
try to solve it before we figure out what the basic headers mean.

Keith