[Mailman-Users] Reply-to options not working

[Stephen J. Turnbull]

Jordan Brown writes:

 > If you have "smart reply" as a separate function, yes.  If you have
 > the typical "Reply" and "Reply All", and the mailing list software
 > sets "Reply-To: <list>", then replying to the author is awkward and
 > error-prone.

Sure, but in this thread we all agree that Reply-To munging is
problematic.  I believe that the widespread availability of "smart
reply" would greatly reduce the incentives for lists to munge.  It
doesn't eliminate the need for reply-author and reply-all, but in my
experience it reduces them dramatically, and basically eliminates the
need for reply-list.

The question I asked, which you misinterpreted completely IMO, and
Grant partially agreed with is "Does an algorithm which 1. gives
overriding precedence to Reply-To, 2. otherwise if List-Post is
present directs it there, and 3. finally falls back to From, seem
likely to DTRT most of the time?"  I know from analyzing my own posts
that for me the answer is yes, and from experience I know that I'm
reasonably good at flexibly using different reply functions to meet
the needs of the moment even though I spend a lot of time using just
one function.

I also think that the evidence is pretty strong that the Reply-To
mungers would like this functionality.  To them, it would seem like
their MUA DTRTs even for non-munging lists.

 > RFC-compliant MUAs

... don't exist.  RFCs about mail do not apply to MUAs, they apply to
the interpretation of the messages.  An MUA can do anything it wants
to.  It's up to the user to decide whether they like the results or
not.  But the user is ultimately responsible for conformance, not the
MUA.

It's true that if a message composed by a borked MUA (eg Thunderbird
in some of its recent incarnations, or whatever the MUA is that
encourages people to add unneeded Reply-Tos) is interpreted according
to RFC 5322 et al, the result is often surprising to the user.  But
you shouldn't assume that proposed features are going to be added in a
borked way.  (Statistically, of course, some *will* be borked on
introduction.  That just means we need to be ready to fix them!)

Here my main question is whether for *many* users "smart reply" would
"DTRT" enough to streamline their UI and reduce mistaken addressing.
It should be an additional option, not replacing any of the now-
traditional features (reply to author, reply to all, reply to list).
(However, in my experience it completely replaces reply to list.)  If
it's not close enough to what a large number of users want to be the
recommended binding for the "obvious" gestures for "generic" reply, I
don't think the additional variety (complexity) in UI configuration is
worth it.

 > So for the general case where you might have gotten a message
 > directly, and through list A, and through list B, the result is
 > random unless you pay careful attention to how you got this
 > particular copy of the message.

Yes and no (I partly disagree with Mark here).  It's definitely
deterministic, and *not* random, but to users it may seem arbitrary.
This can be mitigated in many cases by list owner coordination and
subscriber setup.  So, for example, a subscriber who sorts different
lists' traffic into different folders is likely to be aware which list
it will go to (the List-Id and List-Post fields will be consistent
throughout that folder).  If it's a group of related lists with
different topics (such as mailman3-users, mailman-users, and
mailman-developers) there's likely to be strong social strictures
against cross-posting.  And with hierarchical structure, the owners
can set things up to DTRT.  Examples:

Announce list: does not set List-Post
    Users list: does set List-Post (cross-posting discouraged)
    Dev list: does set List-Post (cross-posting discouraged)

The announce list being gated to users and devs respectively
"obviously" should only get posts from approved sources.  This DTRTs.

Advisees: does not set List-Post
    Undergrads: does set List-Post (cross-posting discouraged)
    Grads: does set List-Post (cross-posting discouraged)
        Graduates 2018: does not set List-Post
        Graduates 2019: does not set List-Post
        Graduates 2020: does not set List-Post

The "all my advisees" list functions as an announce list (when I'm out
of town, all-hands meetings and university deadlines announcements,
etc).  The undergrads and grads are socially distinct and occasionally
have discussions among themselves.  Rarely the 1st, 2d, 3d year grad
students have discussions specific to that year, but mostly discussion
relates to seminar presentations and general research methods, so it's
not that inconvenient to have no List-Post for individual classes.
It's not perfect DWIM, but it's pretty close.  I don't often see this
as being a problem, and lists that set Reply-To would surely make it
worse.

So overall, I don't see this as likely to be a big problem in practice.

 > It sounds like neither of us want the list to set "Reply-To:
 > <list>".

I think that's pretty general feeling among list management software
developers.  Users and list owners, on the other hand, frequently
disagree.  I believe that is due to a deficiency of MUAs, to wit, not
offering a "smart reply" function and not encouraging its use by less
sophisticated users.

 > You want a "smart reply" button that sends to Reply-To, List-Post, or
 > From, in that order.  (Right?)

I would (and do, where available) use it, but I really want it as an
option, even default, for the crowd of users who expect Reply-To to be
set to the list with current MUAs.

 > I wouldn't use your "smart reply" button, because I think it does the
 > wrong thing for mailing lists,

I don't understand why you think that.  So far you have consistently
responded to this thread on-list AFAICS, and everybody in this thread
got here by reading it on the mailing list (all first responded to a
mailing list post, not to one where they were personally addressed).

I think that is by far the majority case.  It is *quite* unusual for
non-subscribers to be explictly addressed in my experience (see below
for an exception), and (for spam control reasons) it is very common
for lists to make it difficult for non-subscribers to post.  It is not
unreasonable in this day and age to assume that all posters are
subscribers, and that all explicitly named addressees got there
because of "reply all" rather than because somebody added a
non-subscriber.  On this list, and many like it, I personally would
rarely ever need anything other than "smart reply", and it also works
fine for most of my personal mail, which almost always should get a
reply only to author (or Reply-To, which smart reply does).

But I'm interested to hear about use cases where that is *not* the
case.

For example, I participate in one list at work where "smart reply"
would often do the wrong thing, because the subscribers to that list
are fellow faculty and staff who deal with applicants to the grad
program.  I end up splitting between smart-reply (which goes to list
for discussion of committee policy and the like) and reply-all (which
goes to the list so others know I dealt with that case, and to the
non-subscriber to whom I provide information).  However, this decision
*must* be made case-by-case, depending on the nature of the reply:
relying on reply-all (as I do on most list traffic when my MUA doesn't
have "smart reply") would quite frequently do the wrong thing and
require non-list addresses to be edited out.  Even when reply-all is
appropriate, a non-subscriber frequently needs to be added by hand to
coordinate a student-adviser pairing (mostly by the staff, not often
by faculty members).  Reply-author gets used, but quite rarely because
of the nature of committee work.  So even though I can't rely on smart
reply to DWIM, it is nevertheless useful, and fails safe (we rarely
care all that much if any given applicant falls through the cracks,
and most cases where the mail goes to the list but not the applicant
gets caught same day, but internal committee discussion should *not*
go to random non-university people!)

Of course that's just *one* use case.  That's why I ask for others.

 > but if you want to do the wrong thing with your replies, I guess
 > that's up to you.

It's not about *me*.  I have this feature in my own MUA (or will, as
soon as I re-add it to this new upgraded version).  And it does the
*right* thing for me more than 95% of the time.  I still don't
understand what makes you think it would do the wrong thing for *you*,
let alone the wrong thing for *me*, except that you object to it being
labelled "Reply" because you associate that very strongly with "Reply
to Author".

 > My only fear is that in the ongoing simplification (dumbing-down?) 
 > of this stuff, "smart reply" will become the only option.  And,
 > actually, if that happens then I *have* lost the "reply to author"
 > function.

I don't think that level of paranoia is justified.  Sure, some dev
organizations will make that kind of mistake, as we've seen with
Thunderbird.  But all of the MUAs I know that do have specialized
reply-to-list (mutt, Gnus) have very flexible interfaces for binding
UI gestures to functions, and far more available functions than
"one-click" or "one-key" gestures.  If yours doesn't, then yes, you're
at risk that a whim of the developers you could lose essential
functionality.  But that's a problem with your MUA and its dev team,
not with the suggested new functionality.

Steve