Re: Reply to list [was: Different key maps in different dired buffers]

[Bob Proulx]

Eli Zaretskii wrote:
> Glenn Morris wrote:
> > Eli Zaretskii wrote:
> > > Either one is okay.
> > 
> > I'm so sure that reply-to-all is correct that I thought it was
> > official policy for GNU lists.
> 
> You may be right.  I just don't think I've ever seen it written, but
> it could be that I missed something.

I don't think anything is written down.  (But I have only been doing
this for ten years or so and still haven't learned all of the
secrets.)  Because for the GNU project we would need to get rms to
approve it and as far as I know no one has pushed an official policy
through for approval.  Until that happens it can only be an defacto
standard.  But by defacto standard reply-all is normal, unless you
know the poster is on the list already.

Pragmatically the bug reporting lists can't be reply only on the
mailing lists.  Because we don't require bug reporters to subscribe
before posting bug reports.  We don't require bug reporters to
subscribe first.  Doing so would be a burden and would alienate bug
reporters.  Therefore for any random bug report we should always
include the original reporter in the recipient list.  (Unless we know
them and know they are already subscribed.)

Because the bug reporting lists are open and need to include the
original sender this carries over to other lists too.  Like this
help-gnu-emacs list.  Should it require people to be subscribed?  That
is more of a gray area since it isn't a bug reporting list.  But
people writing for help are almost the same as bug reporters.  So by
default help lists get the same treatment as bug lists.

But what about discussion lists?  At some point maintainers of
individual projects may state a policy for their own project.  And so
there isn't a 100% uniform policy across all of the mailing lists.
Some lists.{non,}gnu.org lists are one way and some another way.

Other free libre projects have different policies.  For example on the
Debian lists the official policy is to reply only to the list unless
specifically requested by the poster that they be CC'd.  And many
other projects have that policy too.
 
Generally I reply-all unless I recognize the senders as already being
members of the mailing list.  For example here on this message it is
safe to reply only to the mailing list because all of the recipients
obviously read the message that was posted to the mailing list and so
must be subscribed.  I would hate to annoy them with a second copy.
The only unknown is the original poster who said he wasn't subscribed
and therefore I specifically added as a CC.

Additionally there is the problem of the News to Email gateway.
(Which coincidentally I see had a backlog of old messages push through
today.)  Many people read on the News side of things and post there.
I think for them it would be quite surprising to get an email copy of
something they posted by news.  However this is a mailing list and not
a news group.  People reading the news group have to expect that they
are still participating in a mailing list.

> > I don't understand why people think "recipient might get two copies"
> > is worse than "recipient might get no copies". Especially when the
> > former issue is trivially avoided by Mailman or MUA duplication
> > suppression.
> 
> I fully agree.

I completely disagree.

Mailman duplication suppression is mostly non-functional for this
issue.  And when it does come into play it can cause loss of messages
which is the exact opposite of what you want.  MUA duplication
suppression is extremely complicated in order to have it working
reasonably.  Since discarding the second by message-id is almost
always the wrong thing to do.

The problem with people getting two copies is that it is terribly
annoying to manage.  I will almost always get the direct copy first
before the mailing list copy.  And because of Mailman I never know if
I will actually get the Mailman copy.  I have to wait and see.  That
makes management of all of the messages much more complicated.  Which
is exactly why projects that have a policy of only replying to the
mailing list have that policy.

This is solved with Mail-Followup-To: but since that is only another
defacto standard it isn't implemented by many mailers.  Since most of
the popular mailers don't get basic things like threading right it
isn't surprising that they don't implement more subtle things.

And so we have an imperfect system that we simply have to learn to
live with regardless of the problems.  Even with the problems it is
much better than a web forum.

Bob