[sf-lug] email, Reply-to:, lists, and all that jazz

[Asheesh Laroia]

On Tue, 29 Apr 2008, Michael Paoli wrote:

> Well, I would have made several of the excellent points Rick already
> made, but he beat me to it :-) (Thanks, Rick).
>
> In any case (quite a "thread" it turned out to be), there were still
> several items I'd also consider worthy of mention, that weren't
> directly addressed, or were rather to quite lightly touched upon.
>
> Anyway, some of those points and some examples.  I'll try to cover them
> in what I'd think roughly priority order (relative to my perception of
> their importance, and how much they may have not, or not so much, been
> covered already - at last on this "thread") ... such may not necessarily
> be the most logically flowing of order, however ;-) :
>
> * "User" training/education.
>  Part of this "reply" / "Reply-all" / Reply-to "issue"/"problem" I
>  think can be at least somewhat addressed by user education/training.
>  I think also as email client software continues to get better, that
>  also helps things - e.g. in making it more clear - and consistently
>  clear - to the user - what various actions in the email client do (or
>  at least ought to do) and will do.
>  * "r"eply/"R"eply-all - most any reasonably sane email client provides
>    at least two - and in many cases only two - "reply" flavors of
>    options/choices/commands.  They're commonly called/labeled "reply",
>    and "reply-all" or "reply-to-all".  Note also for common single-key
>    commands, that "r" vs. "R" don't always correlate the same way
>    around (probably much more commonly "r" is "reply" and "R"
>    "Reply-all", but sometimes the reverse is the case).
>  * probably one of the easiest ways to think of the "reply" vs.
>    "Reply-all" is a minimal vs. maximal reply.  Though that may not be
>    strictly true in terms of number of recipients, it tends to hold
>    quite true in number of email addresses the client sends to.
>  * "reply" has the client send to a single email address (though that
>    address might be a forwarder, alias or even a list).  "reply" can
>    also be thought of as generally going to the author/originator of
>    the email (or address they specify that replies be directed to).
>    "reply" is *not* for some intermediary (other than as may be
>    explicitly specified by the author/originator).  (Probably a sucky
>    analogy - but ... magazines in the mail, business reply cards -
>    reply - one expects them in that case to go to the magazine or
>    advertiser, not to be delivered to the postal service itself for
>    having made possible this distribution and delivery of large
>    quantities of magazines, ... and not to the 3rd party printing press
>    or mailing house that printed and sent the magazines, either).  So,
>    ... think "reply" - reply to author (or their designee).
>  * "Reply-All" goes more towards the maximum, sending using To: (and
>    sometimes also Cc:) picking the addresses from Reply-To: (or From:
>    if Reply-To: isn't present or is empty), To: and Cc: headers, and
>    using all those addresses (think maximum/broadest/*all* - or at
>    least the largest set that would be a reasonably sensible thing to
>    do).
> * on email, lists, and replying, etc. in general.
>  * If you want to reply just to the author/originator (or their designee), or m
> ostly so, start with "reply".
>  * If you want to address author/originator plus most or all of the email's rec
> ipients, start with "Reply-all".
>  * See also immediately following point!:
>  * before sending, review, and adjust as appropriate, the To:, Cc:,
>    and/or Bcc: recipients.  Not everything will always work as one
>    expected or anticipated (or even as it should), and merely
>    selecting "reply" or "Reply-All" may not have gotten one the set of
>    recipients one actually wanted - so always review these (I'd
>    recommend looking them over and adjusting them immediately after
>    one does the reply or Reply-all selection, and again before making
>    the send action).  One may also want to pay attention to exactly
>    what recipients are/aren't included (did you include lists you're
>    not even subscribed to and may not be able to post to, or email
>    addresses you don't even recognize or know why they're there or
>    who/what they're for?)  Whom is one addressing and/or Ccing?  Is
>    that reasonably clear from the To:/Cc:(/and possibly also Bcc:)
>    recipients, and if appropriate, salutation and/or closing?  Note
>    also that in general, author/originator of email can set Reply-To:
>    to whatever they want - e.g. if they want, they could set it to the
>    list, or some other list, or president at whitehouse.gov, or whatever
>    ... so, ... check before hitting send :-) !
>  * Before making the send action, triple-check.  For most all intents
>    and purposes, there is no unsend (perhaps another sucky analogy -
>    you've hit send? - you just dropped that letter into a USPS
>    mailbox). [0]


> * Reply-to: munging by list software sucks - hopefully most have figured
>  that out by now :-), but here's a bit more illustration of how it
>  tends to suck.  Example, I often send email, from or as a "role" -
>  i.e. I wear many hats, and may send something under a particular
>  hat/"role".  One of the ways I commonly do that is by setting an
>  appropriate Reply-to: header.  Why?  So that when a recipient
>  replies, the email will go to the appropriate person(s) and/or
>  queue(s) for that role, rather than simply and generally only going
>  to me.  Reply-to: munging breaks that - E.g. I send some BALUG newsy
>  announcement thingy to some spattering of other UGs that are likely
>  interested (e.g. to their talk/discussion lists) - I may set a
>  Reply-To: corresponding to BALUG's publicity feedback (are we getting
>  too much stuff out there, not enough, not the right places, or not
>  the best timing, etc.?) ... with the Reply-to: set, that feedback
>  goes to the folks that work on trying to find that "just right"
>  contents, amount, timing and placement of BALUG's
>  publicity/announcements and communications ... rather than merely
>  going to me.  If the list software munges the Reply-To: to be the
>  list, then when persons read that list, they pick "reply", it not
>  only won't go to those intended by the author of the original
>  message, but I might not even see it at all (at least hypothetically
>  it could be a list I could send to without being subscribed to), and
>  the person responding might be embarrassingly (or worse) surprised
>  that their "reply" went to the list, rather than the author or their
>  designee (think of a newbie having just joined the list, and having
>  never encountered such aberrant standards defiant behavior).

As an alternative, you could just set "From:" to the BALUG role address 
then.  Then you have to decide if you're willing to "lie" in your From: 
address so that your mail is resilient to the intentional corruption (at 
least, as you see it) from other software about to receive and process the 
message.

These days, changing the "From:" is the way people handle multiple roles 
usually - that's why Gmail and others offer a feature by which you can 
change your From.  I'm not saying you must conform to that, though.

In my alpine configuration, I keep "From:" a free-form editable field the 
same way I keep "To:" and "CC:" free-form editable fields.

I do understand the other issues, so you don't need to re-raise them; if 
there are specific responses to this particular response of mine, then by 
all means let 'em flow.

>  Anyway, many other suitable references and points on
>  the topic were covered, but I thought I'd add some much more direct
>  example (like yes, it would in fact impact email I send - even
>  including to this list; and would likewise impact email of *everyone*
>  sending to the list if Reply-To: munging were in effect).

I hope you see that if you munge your own headers in preparation for 
others to munge your headers also, you can avoid this issue.  I'm not 
saying you must do that, or should like to do that.

-- Asheesh.

P.S. On a list without reply-to munging, the CC: line becomes enormous 
when everyone replies-all.  I use a procmail+formail filter to remove 
duplicate inbound messages based on their Message IDs - but if no munging 
+ "use Reply to all" is the most elegant solution, it's still not very 
elegant.  I'm not demanding anyone change their actions, nor claiming that 
what has been said on this thread isn't true, just saying that this is a 
mess either way.

-- 
A day without sunshine is like a day without Anita Bryant.