Date: Fri, 24 Oct 1997 16:55:57 -0700 (PDT)
From: Chris Newman <Chris.Newman@innosoft.com>
Subject: Re: REPLY problem list (draft #1)
In-reply-to: <199710232243.SAA13834@spot.cs.utk.edu>
To: Keith Moore <moore@cs.utk.edu>
Cc: drums@cs.utk.edu
Message-id: <Pine.SOL.3.95.971024155158.8818K-100000@eleanor.innosoft.com>
MIME-version: 1.0
Content-type: TEXT/PLAIN; charset=US-ASCII
Originator-Info: login-id=chris; server=thor.innosoft.com

I like your problem list.  I think it needs to be analized in conjunction
with uses of reply to.  Current uses of the reply-to header are:
 
(A) The originator wishes to express address to use instead of the from
address (e.g., redirecting personal replies to secretary or other account)
 
(B) Same as (A), but the originator prefers a personal reply (e.g.,
soliciting private input on a mailing list).
 
(C) The originator wishes to express addresses for group replies to use   
instead of the from/to/cc addresses (e.g., prevent duplicates).
 
(D) Same as (C), but the originator prefers a group reply (e.g., announce
discussion to multiple lists, direct replies to one).
 
(E) A mailing list prefers that responses go to the mailing list and
wishes limited-function UIs to be presented with the list address rather  
than the from address.

(F) The mailing list prefers that subscribers not receive duplicate copies
of list replies.

(note that the common UI model ignores the distinction between A/C and B/D 
and another UI models treat B/D the same and ignores A/C.  I don't think
we can legislate either in a standard although we can encourage richer
UIs.)

So let's see what your problems list tells us about the reply-to uses:

On Thu, 23 Oct 1997, Keith Moore wrote:
> 1.  Given the vast number of MUAs that implement "reply all" but
>     behave as "Reply-To replaces only From",  there's no way for
>     the author of a message to say "please reply only to this 
>     address". (except perhaps to use Bcc for all recipients)

This suggests UAs need to be able to distinguish (B)/(D) from (A).

> 2.  If a mailing list mungs an existing Reply-To field, there's no 
>     way for the author to request a different reply behavior
>     (as in "please reply to XXX, not to wherever the list wants 
>     you to reply") 
> 3.  If a mailing list mungs an existing Reply-To field, there's
>     no way for the recipient to honor the author's requested
>     reply-to behavior.
> 4.  If a message is sent to multiple recipients, one or more of which
>     is a mailing list, and one of those mailing lists adds a Reply-To 
>     field, members of that mailing list have difficulty replying to
>     all recipients of the original message.  This thwarts cross-list
>     discussions, which are sometimes desirable.
> 11. If a message is sent to a mailing list that adds or mungs
>     reply-to, and the author of that message is not on the mailing list,
>     any replies that go just to Reply-To won't go to the author.
>     Even replies that go to Reply-To+To+CC won't go to the author
>     of the subject message unless the author has included himself
>     in the To or Cc list.

These suggest that (E)/(F) need to be independent from (A)-(D).

> 5.  If a mailing list leaves an existing Reply-To field in place,
>     but supplies a Reply-To field if one is missing, there's no
>     way for the recipient's UA to know whether the field was added
>     by the list or not.  The results can be surprising.

This suggests we need to be able to distinguish (E)/(F) from (A)-(D).

> 6.  If a message contains a Reply-To field, but the person composing
>     the reply isn't aware that Reply-To is being used, the reply may
>     go to somewhere besides where the reply author thought it should
>     go. The result may be surprising or embarassing (as when a
>     personal reply gets sent to a list), or the reply author may
>     think he has sent the message to a list when it has really only
>     gone to one particular recipient.

This is a combination of a UI issue and the need to distinguish (A)/(B)
from (C)-(F).

> 7.  A few MUAs implement only one kind of reply -- which uses 
>     Reply-to if present and From otherwise.  Users of such MUAs
>     have difficulty participating in group discussions on mailing 
>     lists (unless the list mungs Reply-To).

This indicates option (E) is necessary.

> 8.  If a message is sent to a mailing list, and someone does a 
>     "reply all" to that message, the author often gets two copies -
>     one sent to the From address and another sent to the list
>     address (which presumably appeared in the To or Cc field
>     of the subject message).
> 9.  If a user is subscribed to multiple mailing lists, and a message
>     is sent to more than one of those mailing lists, the user will
>     get multiple copies of the message.  This might not be so bad,
>     but the same user will also get duplicate copies of the message 
>     every time someone does "reply to all" to the original messages 
>     or any of the replies to that message.
> 12. If someone responds to a list message and CCs themselves in
>     the response, everyone who does a "reply all" to that message
>     will also Cc that person...even if he's on one of the lists.

These indicate that (F) is insufficient and users need to know which
lists they're subscribed to and express that information accordingly in
headers -- probably via (C)/(D).
 
> 10. On many MUAs, if the sender specified Reply-To, the recipient is 
>     stuck with that - he cannot easily edit it to reply to From
>     or From+To+CC (especially if he cannot see the original headers
>     while he's editing the reply recipient list).  The author of
>     the reply thus sometimes has difficulty sending the reply to where
>     it needs to go.

This is partly a UA flexability problem which will be decided by customer
selection and partly that restrictive UA behavior is not appropriate for
many of the cases in (A)-(F).

>     Alternatively, if multiple recipients (often lists) get CCed 
>     on a message, every "reply all" goes to all of those recipients, 
>     even if the discussion topic wanders and it's no longer
>     appropriate for many of those recipients.

This suggests large recipient lists need to be editable.

> 13. Many people have multiple mailboxes, or regularly send mail from
>     some account at which they don't want to receive mail.  If they
>     use Reply-To to point to their preferred mailbox, it will prevent
>     some recipients from doing "reply to all".  Also, since there's
>     no indication as to *why* the author used reply-to (it could be
>     "please don't mail to me at my From address" or it could be  
>     "please tell my secretary, not me, whether you'll be at the 
>     meeting"), it's easy for a reply to go to the wrong place.

This suggests (A) needs to be distinguished from (C)/(D).  Whether (A)
should be independent of (C)/(D) is controversial.

		- Chris