OT: Reply-To

[Christian Rose <menthos menthos com>]

Ken Fox wrote:
> > Most lists have Reply-To: <list> by default, and with those lists
> > there are seldom problems. People don't send duplicate messages and
> > don't accidentally only reply to the sender.
> 
> You *still* don't get it?!

If you mean accepting a broken behavior, no I won't. You haven't
convinced me that that broken behavior should be something better yet,
and I doubt you will.


> Most lists I read (Perl 5/6, Linux kernel, Apache mod_perl) do not munge
> the reply-to headers. I have no idea about your statement about "most"
> lists, but I suspect that you're hand-waving.

What? Let's see. These are some of the other lists I'm subscribed to:

list					uses what?
-------------------------------------------------------------------------
galeon-devel@lists.sourceforge.net	Reply-To (I suspect that all
					Sourceforge lists do that by default)
gnome-i18n@gnome.org			No Reply-To
i18n-list@redhat.com			Reply-To (I subscribe to a lot of
					other RH mailing lists too, each of
					them using Reply-To)
swe_doc_listan@listserver.nu		Reply-To (This is the Swedish LDP
					translation list)
sv@li.org				Reply-To (Swedish Translation Project
					list)

(yes, I'm a lot into translation :-).
You see that only the gnome mailing lists don't have the Reply-To
setting.


> Read the following summary for why reply-to is best left alone:
> 
>   http://www.unicom.com/pw/reply-to-harmful.html

Oh, I've read it now. Thanks for the link. I'll comment it here:
[quotes from the link above. In my responses, I'm directing my replies
to the author of that text.]

> Reply-To munging does not benefit the user with a reasonable mailer.
> People want to munge Reply-To headers to make ``reply back to the list''
> easy. But it already is easy. Reasonable mail programs have two separate
> ``reply'' commands: one that replies directly to the author of a message,
> and another that replies to the author plus all of the list recipients.
> Even the lowly Berkeley Mail command has had this for about a decade.

Yes. All mailers have a "Reply" and a "Reply-All" function. The problem
with having to use Reply-All is that it will send duplicate mails to the
recipient (he will recieve mails BOTH from the list and you directly),
unless you edit the sender field manually to point it only to the list.
That's a big problem, because a lot of people don't do that, either
because of laziness or that they forget it.


> Any reasonable, modern mailer provides this feature. I prefer the Elm
> mailer. It has separate ``r)eply'' and ``g)roup-reply'' commands.
[snip more wonderful praising of Elm]
> Whichever mailer you choose, please read the fine manual that comes with
> it. Unless you are stuck with some decrepit mail system, I bet you'll
> find it has a similar feature. If so, you easily can choose to direct your
> responses either to the original author or the entire list.

That's what's wrong. It isn't replying to "the original author or the
entire list" in most mailers. It's replying to "the original author AND
the entire list" in most mailers if you're using the Reply-All feature,
and it results in the original author getting spammed with duplicate
identical messages, both from the list and you directly.

The mailer doesn't have built-in logic so that it knows that replying to
the list also replies to the original author. It simply couldn't have,
unless each mailer has a list of all recipients to a list. That's why
this isn't a mailer problem at all, but a problem that should be solved
on the list server side.
It's a server problem, not a client one, and that's why this argument
fails.


> If you use a reasonable mailer, Reply-To munging does not provide any
> new functionality. It, in fact, decreases functionality. Reply-To munging
> destroys the ``reply-to-author'' capability. Munging makes this command
> act effectively the same as the ``reply-to-group'' function.

No. A Reply-To mail header will NOT make replying act the same as
Reply-To replying. Reply-To replying makes sure that the reply will only
go to the list, and will thus ensure that no duplicate messages get
sent, as with using Reply-All. That's a big difference.


> We haven't added anything new, we've only taken away. Reply-To munging
> is not merely benign, it is harmful. It renders a useful mail capability
> inoperative.

This is pure BS. Adding Reply-To is a feature, a feature that ensures
that replies will go to the list, and to the list only, unless other
recipients are added. This way, there WON'T be unnecessary duplicate
mailings that only clutters people's inboxes.

Discussion mailing lists are designed for group discussion, and thus, by
design, most of your replies will be to the list. Replies only to the
original author will be exceptions. In that case, you reply to the
original author only. It's simple.
In fact, I know that at least pine asks you the question when replying:
"Use Reply-To list-adress instead of Reply-To Sender?". That feature
makes it easy when replaying, no matter the circumstances.

This is a feature that CAN be built into all mailers, in contrast to a
feature that would keep track of everybody subscribed to a list so that
there won't be duplicate messages sent from the mailer, which isn't
realistic.


> Some administrators justify Reply-To munging by saying, ``All responses
> should go directly to the list anyway.'' This is arrogant. You should
> allow me to decide exactly how I wish to respond to a message. If I feel
> a public response is justified, I'll hit the ``g'' key and tell Elm to do
> a group-reply. If I believe a private response is more appropriate, I'll
> use ``r'' to send one. Please allow me the freedom to decide how to handle
> a message. 

You have a freedom to send to whomever you want to. If your wonderful
mailer Elm had a function for selecting if you'd like to use Reply-To
sender instead of Reply-To list, you would have no problems with
Reply-To at all.

The freedom to send to whomever you wan't to mustn't be misused by
spamming people with multiple messages. That's why Reply-To exist. It
insures that, by default, the original author won't get spammed with
multiple identical replies from the same sender, unless the sender
modifies the To: fields directly to do that said spamming.


> It may be impossible to reply to the author of a message once the Reply-To
> header is munged. The Reply-To header was not invented on a whim. It is
> there for the sender of a mail message to use. If you stomp on this header,
> you can lose important information. 
> 
> There are good reasons why the sender might insert a Reply-To header. The
> sender might not be the original author of the message (the name that
> appears in the From header). If responses should return to the sender and
> not the original author, then the sender will insert a Reply-To header.
> Or, maybe the sender added a Reply-To because he or she cannot receive
> email at the account from which the message was sent. There are many good
> reasons to place a Reply-To header into a mailing list message. 
> 
> If the Reply-To is munged by the mailing list, the value provided by the
> original sender is lost. Reply-To munging can make it impossible to reach
> the sender of a message.

But there exists already a field that takes care of that (and the
mailing list software should be aware of)!
I proudly present to you the RESENT-REPLY-TO header, described in RFC
822, quoted at the end of this letter.
Boom goes that argument.


> There are, unfortunately, poorly implemented mail programs that lack
> separate reply-to-author and reply-to-group functions. A user saddled with
> such a brain-dead mailer can benefit from Reply-To munging. It makes it
> easier for him or her to send responses directly to the list.
> 
> This change, however, penalizes the conscientious person that uses a
> reasonable mailer. This is a poor trade-off. As Internet list
> administrators, we should encourage people to run reasonable software.
> If a few people need to type in a full reply address so that everybody
> else can use all the features of their mailer, I say, ``Fine!'' We should
> not penalize the conscientious to coddle those who run brain-dead software.

This is funny because:
* Most mailers DO have both Reply and Reply-All (or group reply as you
call it)
* You call all other software brain-dead. Well, I would say that Elm is
brain-dead in this case, as it appearantly doesn't let you choose
between using the sender address or the Reply-To address in an easy way.


> Compare and contrast: the work required for me (or any other Elm user)
> to reply on lists that do and don't employ Reply-To munging.
[goes on to explain the horror of extra commands]

Well, if Elm had let you choose between using the Reply-To address or
the sender address...


> When I hit the ``r'' key in Elm, it sends a response to the author of a
> message. When you munge the Reply-To header you change this action so
> that it does something entirely different from what I expect. This
> creates specialized behavior for your mailing list, which increases the
> potential for surprise. I'm not schooled in the science of human
> factors, but I suspect surprise is not an element of a robust user
> interface.

Okay, so you take an UI argument. That's interesting, because your
conclusion is wrong in my opinion, and I guess others would agree. The
reason for this is simple:

When you send a message to a mailing list, it already behaves
differently!

When you send a message to a mailing list address, it gets sent to a lot
of addresses. This is a mailing list fundamental, and one that users
learn.

When you reply to a message from a mailing list, it gets sent to a lot
of addresses (when using Reply-To in the list). That's fairly
consistent, and an easy to grasp concept. It behaves exactly like you
would expect from a mailing list.


> Private messages frequently are broadcast across lists that do Reply-To
> munging. That's an empirical fact. It's what happens when you violate the
> principle of least surprise. 

In my experience, that's simply wrong. In my whole experience using
Reply-To mailing lists, there have only been a handful accidents where
people accidentally sent replies to a single person to the whole list.

On the other hand, using the gnome-gui list for only a month, there have
been accidents several times a week where people expected the reply to
go to the list, and not just sending it to me.


> Consider the damage when things go awry. If you do not munge the Reply-To
> header and a list subscriber accidentally sends a response via private
> email instead of to the list, he or she has to follow up with a message
> that says, ``Ooops! I meant to send that to the list. Could you please
> forward a copy for me.'' That's a hassle, and it happens from time to time.
> 
> What happens, however, when a person mistakenly broadcasts a private
> message to the entire list? If the message is a complaint about the
> personal hygiene of sender's boss, or the sex life of his or her roommate,
> a simple ``Ooops!'' won't cut it. About all you can do is send a followup
> with lots of retroactive smileys (weak). Or say your cat was dancing on
> the keyboard (better). Or start reading the classifieds for a new
> job/roommate/set of teeth (most likely). 

But that the reply goes to the whole list is what people _expect_! If
you send a message to a list, it goes to all subscribers. If you get a
response on that, you'd expect that it had gone to all subscribers. If
you resond to that, you'd expect it to go to all subscribers. That is
the concept of mailing lists. Discussions go to a lot of subscribers,
unless you modify, either by not sending it to the list in the first
place (by typing in an other address), or modify it in the second place
(by typing in an other address) in your following replies.

Since people expect that mailing lists go to a lot of people, they take
care when typing information, and do not include sensitive or insultive
content when posting to a mailing list. Likewise, they take the same
care when replying to a list.

I've _never_ seen people accidentally replying with sensitive or
offensive material to a mailing list. People actually take care when
writing to mailing lists, unless they're completely screwed and don't
realize the potential consequences to begin with.


> If you are not convinced yet, then allow me one final plea. I contribute
> to the Elm mailer development team. I get to see a lot of the wants and
> requests from the user community. Guess what feature more and more people
> are asking for? A third reply command -- one that ignores any existing
> Reply-To header! Want to guess why people are asking for it?

Yes, they want an easy way to control the Reply-To behavior that has
increasingly become a standard on mailing lists for very good reasons.
By denying your users this feature, you're not making them a favor.
And by claiming the evilness of Reply-To by citing the behaviors of Elm
which appearantly doesn't have this feature, you're indeed making a
strange argument.


> In case you are wondering, yes, I once thought Reply-To munging was a
> nifty idea. I got better though.
> 
> When I started running email lists, I munged 'em all. One day I
> accidentally sent a private, personal reply out over one of my own damn
> lists. If the list owner can't remember how to use the list properly, no
> way will the subscribers be able to sort it out. I stopped munging the
> very next day.

Ah, here it comes. The author got bitten once by his own mistake, and
has since then been on an crucade for fighting the evilness of Reply-To.

Well, you know what? I get bitten several times a week, either by people
sending me multiple replies, or people sending me replies that were
intended for the whole list, and to which I have to reply to them and
explain the error to them, unless they didn't catch their error
themselves.

I can't control other people's mistakes. However, I can control my own
behavior and my own mistakes. If I erroneously send a message to a whole
list, that was my own mistake. I should have taken more care.

If other people send me multiple messages, it's their error. I can't
control that. I have to politely ask them not to do that. Some accept
that, some plain refuse to do that and insist that sending people
multiple mails is the correct behavior.

Personally, I'd rather be responsible for my own mistakes, thank you,
than having to accept other people's mistakes all the time.


> On the whole, it has worked out quite well. Yes, on occasion somebody
> mistakenly responds directly to the author of a message when they wanted
> to reply to the group.

On occasion? It happens to me all the time. People don't seem to care
about it, or forget the no-reply-to behavior. That's a real problem.


> Most folks, however, seem to catch on pretty fast to how it works, and
> seem to appreciate the flexibility. Moreover, private responses mistakenly
> sent to the entire list have become an almost unheard-of event. 

To the price of insane mail-spamming and user frustration, yes.


> If you think you are doing your subscribers a service by munging Reply-To
> headers, you are kidding yourself. You are making your subscribers
> miserable. 

Excuse me? Lists WITHOUT Reply-To are the ones who make MY life
miserable. I get multiple mails all the time, and I can't control that
behavior. All thanks to some list admins that are refusing to set
Reply-To, by claiming your broken "reasoning".

On the other hand, I never have such problems with most of the lists
that use Reply-To. Zip. Nada. Zilch.


> Some list administrators, even after reading all this, seem to say, ``Oh,
> it's not that bad. Besides, my subscribers like it!'' If they do, it's
> probably because they haven't bothered to learn to use the
> ``reply-to-group'' feature of their mailer. Instead of going through all
> the trouble of making your list gateway scribble on email headers, how
> about making an effort to educate your subscribers? 

Again, the author shows that he doesn't understand the difference
between Reply-All (aka Reply-To-Group) and Reply-To. They're not
equivalent, and one of them is not a replacement for the other.

And "educating subscribers" to compensate for a broken default behavior?
Please.


So, that was that page. I'm going to reply to Ken's additional statement
now:

> If you don't want duplicates, you can either set your own personal
> reply-to header when you send to the list, OR filter out duplicates when
> mail is delivered to you (procmail or similar), OR get a smarter IMAP
> agent, OR remind people in every message you send not to send directly
> to you, OR come up with a new X- header that hints mailers not to send
> directly to you in a list reply.

_You_ don't seem to get it, do you? Your "solution" is plain ridicolous.
The reasons why:

1) Most people don't have control over their mail servers
2) Most people don't have complete control over their mail clients
(e.g. thay can't hack it to add an extra header)
3) Most people have to rely to the mail filtering features of their mail
clients, which is often very limited (most mail mail client filtering
only sorts mail, it doesn't try to find duplicates). Not evryone runs
procmail or has the possibility to do so
4) You can't ask every mailing list user that accidentally sends you
duplicate messages multiple times to stop doing so, unless you want to
sit and work with your mail full-time
5) Some people disregard your polite request (like you do with my
request,  suggesting that there is something wrong with my request of
you not to send me duplicate mails)

But I think that the most erroneous thing about your "solutions" is that
thay are _client-side_ hacks for a _server-side_ problem.

Should we base mailing lists on the principle of client-side filtering,
so that people will have to have filters for duplicate mails? Isn't this
that ultimately should be fixed on the server side instead? Reply-To
does just that, and isn't dependant on ALL mailing list users having to
hack their software. Instead, it's a simple config option in the mailing
list server.


Finally, I'll add some good reference:


        August 13, 1982              - 21 -                      RFC
#822

 
     Standard for ARPA Internet Text Messages

     4.4.3.  REPLY-TO / RESENT-REPLY-TO

        This field provides a general  mechanism  for  indicating  any
        mailbox(es)  to which responses are to be sent.  Three typical
        uses for this feature can  be  distinguished.   In  the  first
        case,  the  author(s) may not have regular machine-based mail-
        boxes and therefore wish(es) to indicate an alternate  machine
        address.   In  the  second case, an author may wish additional
        persons to be made aware of, or responsible for,  replies.   A
        somewhat  different  use  may be of some help to "text message
        teleconferencing" groups equipped with automatic  distribution
        services:   include the address of that service in the "Reply-
        To" field of all messages  submitted  to  the  teleconference;
        then  participants  can  "reply"  to conference submissions to
        guarantee the correct distribution of any submission of  their
        own.

        Note:  The "Return-Path" field is added by the mail  transport
               service,  at the time of final deliver.  It is intended
               to identify a path back to the orginator  of  the  mes-
               sage.   The  "Reply-To"  field  is added by the message
               originator and is intended to direct replies.

     4.4.4.  AUTOMATIC USE OF FROM / SENDER / REPLY-TO

        For systems which automatically  generate  address  lists  for
        replies to messages, the following recommendations are made:

            o   The "Sender" field mailbox should be sent  notices  of
                any  problems in transport or delivery of the original
                messages.  If there is no  "Sender"  field,  then  the
                "From" field mailbox should be used.

            o   The  "Sender"  field  mailbox  should  NEVER  be  used
                automatically, in a recipient's reply message.

            o   If the "Reply-To" field exists, then the reply  should
                go to the addresses indicated in that field and not to
                the address(es) indicated in the "From" field.


Quite interesting read, isn't it? Makes one wonder if supporters of the
no-reply-to behavior even knows what they're talking about when they say
that Reply-To is entirely non-standard in discussion lists. Quote:

"A somewhat different use may be of some help to 'text message
teleconferencing' groups equipped with automatic distribution services
include the address of that service in the 'Reply-To' field of all
messages submitted  to the teleconference; then participants can 'reply'
to conference submissions to guarantee the correct distribution of any
submission of their own."

Couldn't have said that better myself.


Over and Out, hoping for a change.
Christian



#######################################################################
Christian Rose
http://www.menthos.com                    	    menthos@menthos.com
#######################################################################