Mutt group reply

Mutt group reply

[Solar Designer]

Hi,

I noticed that recent replies by Isaac Dunham and Szabolcs Nagy to
messages by Brent Cook went to everyone but Brent.  This is typically
caused by a side-effect of the list setting Reply-To to point to the
list, combined with the way the group reply feature is implemented in
Mutt.  When you hit "g" in Mutt (for group reply), you're prompted:

Reply to musl@lists.openwall.com? ([yes]/no):

If you say "y", Mutt assumes that the Reply-To is the sender's address,
so it excludes the address in From from those it sends the message to.

What you need to do is answer "n" to that question.  Then it actually
replies to all, including to the list.

An alternative is to reconfigure the list so that it doesn't set the
Reply-To header, but this may result in many replies being inadvertently
sent off-list.  I think it's better for Mutt users to adopt a habit to
answer that question with "n".

Arguably, this Mutt feature/default is broken and needs to be changed.

Alexander

Re: Mutt group reply

[Szabolcs Nagy]

* Solar Designer <solar@openwall.com> [2014-07-13 20:34:21 +0400]:
> Mutt.  When you hit "g" in Mutt (for group reply), you're prompted:
> 
> Reply to musl@lists.openwall.com? ([yes]/no):
> 
> If you say "y", Mutt assumes that the Reply-To is the sender's address,
> so it excludes the address in From from those it sends the message to.
> 
> What you need to do is answer "n" to that question.  Then it actually
> replies to all, including to the list.

thanks

will do this from now on

Re: Mutt group reply

[Laurent Bercot]

On 13/07/2014 17:34, Solar Designer wrote:

> An alternative is to reconfigure the list so that it doesn't set the
> Reply-To header, but this may result in many replies being inadvertently
> sent off-list.  I think it's better for Mutt users to adopt a habit to
> answer that question with "n".

  Even better: configure Mutt to use Mail-Followup-To.
  http://cr.yp.to/proto/replyto.html

  Mutt supports it, it just needs to be configured via an option. I can't
remember the details because it's been a long time, but back when I was
using Mutt, I did that for all the mailing-lists I was subscribed to, and
it worked flawlessly.

-- 
  Laurent

Re: Mutt group reply

[Solar Designer]

On Sun, Jul 13, 2014 at 07:59:04PM +0100, Laurent Bercot wrote:
> On 13/07/2014 17:34, Solar Designer wrote:
> 
> >An alternative is to reconfigure the list so that it doesn't set the
> >Reply-To header, but this may result in many replies being inadvertently
> >sent off-list.  I think it's better for Mutt users to adopt a habit to
> >answer that question with "n".
> 
>  Even better: configure Mutt to use Mail-Followup-To.
>  http://cr.yp.to/proto/replyto.html
> 
>  Mutt supports it, it just needs to be configured via an option. I can't
> remember the details because it's been a long time, but back when I was
> using Mutt, I did that for all the mailing-lists I was subscribed to, and
> it worked flawlessly.

This suggestion keeps coming up, but I think it's a solution to a
different problem.  The mailing list sets Reply-To to keep discussions
on the list regardless of what MUA people are using.  When the mailing
list is configured that way, Mutt exhibits the behavior with group
replies that I have mentioned, and the workaround is either to answer
"n" to the question or to set the reply_to option differently:

       reply_to
              Type: quadoption
              Default: ask-yes

              If  set, Mutt will ask you if you want to use the address listed
              in the Reply-To: header field when replying to  a  message.   If
              you answer no, it will use the address in the From: header field
              instead.  This option is useful for reading a mailing list  that
              sets the Reply-To: header field to the list address and you want
              to send a private message to the author of a message.

This problem and the workaround apply even if the mailing list in
question is configured in Mutt as such, in which case Mutt would set
Mail-Followup-To on your messages, but that's irrelevant.

If I understand correctly, Mail-Followup-To reduces occurrences of
duplicate replies to you, but it does not prevent missed group replies
from you to off-list message senders when the list sets Reply-To.

Alexander

Re: Mutt group reply

[Rich Felker]

On Mon, Jul 14, 2014 at 12:58:59AM +0400, Solar Designer wrote:
> On Sun, Jul 13, 2014 at 07:59:04PM +0100, Laurent Bercot wrote:
> > On 13/07/2014 17:34, Solar Designer wrote:
> > 
> > >An alternative is to reconfigure the list so that it doesn't set the
> > >Reply-To header, but this may result in many replies being inadvertently
> > >sent off-list.  I think it's better for Mutt users to adopt a habit to
> > >answer that question with "n".

Thanks! I had been looking for a solution to this issue for a long
time but didn't bother to really look into it.

> >  Even better: configure Mutt to use Mail-Followup-To.
> >  http://cr.yp.to/proto/replyto.html
> > 
> >  Mutt supports it, it just needs to be configured via an option. I can't
> > remember the details because it's been a long time, but back when I was
> > using Mutt, I did that for all the mailing-lists I was subscribed to, and
> > it worked flawlessly.
> 
> This suggestion keeps coming up, but I think it's a solution to a
> different problem.  The mailing list sets Reply-To to keep discussions
> on the list regardless of what MUA people are using.  When the mailing
> list is configured that way, Mutt exhibits the behavior with group
> replies that I have mentioned, and the workaround is either to answer
> "n" to the question or to set the reply_to option differently:
> 
>        reply_to
>               Type: quadoption
>               Default: ask-yes
> 
>               If  set, Mutt will ask you if you want to use the address listed
>               in the Reply-To: header field when replying to  a  message.   If
>               you answer no, it will use the address in the From: header field
>               instead.  This option is useful for reading a mailing list  that
>               sets the Reply-To: header field to the list address and you want
>               to send a private message to the author of a message.

What's really needed is for mutt to have a second variable like
reply_to but that's used for reply-to-all rather than plain reply. I
think this is something we could propose upstream, and probably easy
to patch in. Or maybe there's already a way to do it with hooks.

Rich

Re: Mutt group reply

[Rich Felker]

On Sun, Jul 13, 2014 at 11:51:44PM -0400, Rich Felker wrote:
> > >  Even better: configure Mutt to use Mail-Followup-To.
> > >  http://cr.yp.to/proto/replyto.html
> > > 
> > >  Mutt supports it, it just needs to be configured via an option. I can't
> > > remember the details because it's been a long time, but back when I was
> > > using Mutt, I did that for all the mailing-lists I was subscribed to, and
> > > it worked flawlessly.
> > 
> > This suggestion keeps coming up, but I think it's a solution to a
> > different problem.  The mailing list sets Reply-To to keep discussions
> > on the list regardless of what MUA people are using.  When the mailing
> > list is configured that way, Mutt exhibits the behavior with group
> > replies that I have mentioned, and the workaround is either to answer
> > "n" to the question or to set the reply_to option differently:
> > 
> >        reply_to
> >               Type: quadoption
> >               Default: ask-yes
> > 
> >               If  set, Mutt will ask you if you want to use the address listed
> >               in the Reply-To: header field when replying to  a  message.   If
> >               you answer no, it will use the address in the From: header field
> >               instead.  This option is useful for reading a mailing list  that
> >               sets the Reply-To: header field to the list address and you want
> >               to send a private message to the author of a message.
> 
> What's really needed is for mutt to have a second variable like
> reply_to but that's used for reply-to-all rather than plain reply. I
> think this is something we could propose upstream, and probably easy
> to patch in. Or maybe there's already a way to do it with hooks.

I found a new config that's working really well for this issue:

ignore_list_reply_to=yes
macro index r <list-reply><reply>

The ignore_list_reply_to=yes option fixes the breakage in 'g'
(omitting the sender of the message being replied to) but breaks 'r'.

The second line is a huge hack: it replaces the 'r' binding by a macro
which first performs list-reply (which errors out of the message being
replied to is not in a recognized list) and then performs reply (which
magically gets lost if the list-reply command path was already taken
successfully. So pressing 'r' on a mailing list message always replies
to the list, regardless of whether the list set a reply-to header
(this will save me from accidentally replying off-list on the busybox
list all the time) and pressing 'r' on a non-list message does an
ordinary reply.

Rich

Re: Mutt group reply

[Rob Landley]

On 07/13/14 22:51, Rich Felker wrote:
> On Mon, Jul 14, 2014 at 12:58:59AM +0400, Solar Designer wrote:
>> On Sun, Jul 13, 2014 at 07:59:04PM +0100, Laurent Bercot wrote:
>>> On 13/07/2014 17:34, Solar Designer wrote:
>>>
>>>> An alternative is to reconfigure the list so that it doesn't set the
>>>> Reply-To header, but this may result in many replies being inadvertently
>>>> sent off-list.  I think it's better for Mutt users to adopt a habit to
>>>> answer that question with "n".
> 
> Thanks! I had been looking for a solution to this issue for a long
> time but didn't bother to really look into it.

Thunderbird's "reply all" will _only_ reply to the reply-to, and I wind
up manually copying in individual email addresses to cc: when I bother.
(Yes, it has a 'reply list' button, but the reply-to header overrides
the difference.)

So it breaks other mail clients too. Largely because reply-to seems to
be used so seldom, and thus isn't particularly debugged. (This is the
only list I've followed in the past 5 years at least that uses reply-to.)

>>>  Mutt supports it, it just needs to be configured via an option. I can't
>>> remember the details because it's been a long time, but back when I was
>>> using Mutt, I did that for all the mailing-lists I was subscribed to, and
>>> it worked flawlessly.
>>
>> This suggestion keeps coming up, but I think it's a solution to a
>> different problem.  The mailing list sets Reply-To to keep discussions
>> on the list regardless of what MUA people are using.  When the mailing
>> list is configured that way, Mutt exhibits the behavior with group
>> replies that I have mentioned, and the workaround is either to answer
>> "n" to the question or to set the reply_to option differently:

My solution is mostly just not to participate on the broken mailing list.

Rob

Re: Mutt group reply

[Rich Felker]

On Tue, Jul 15, 2014 at 09:27:09AM -0500, Rob Landley wrote:
> On 07/13/14 22:51, Rich Felker wrote:
> > On Mon, Jul 14, 2014 at 12:58:59AM +0400, Solar Designer wrote:
> >> On Sun, Jul 13, 2014 at 07:59:04PM +0100, Laurent Bercot wrote:
> >>> On 13/07/2014 17:34, Solar Designer wrote:
> >>>
> >>>> An alternative is to reconfigure the list so that it doesn't set the
> >>>> Reply-To header, but this may result in many replies being inadvertently
> >>>> sent off-list.  I think it's better for Mutt users to adopt a habit to
> >>>> answer that question with "n".
> > 
> > Thanks! I had been looking for a solution to this issue for a long
> > time but didn't bother to really look into it.
> 
> Thunderbird's "reply all" will _only_ reply to the reply-to, and I wind
> up manually copying in individual email addresses to cc: when I bother.
> (Yes, it has a 'reply list' button, but the reply-to header overrides
> the difference.)

Reply-to headers should not override the 'reply to all' feature in a
mail client. If they do, this is a bug. What use is 'reply to all' if
it behaves the same as plain 'reply'?

> So it breaks other mail clients too. Largely because reply-to seems to
> be used so seldom, and thus isn't particularly debugged. (This is the
> only list I've followed in the past 5 years at least that uses reply-to.)

In lists I've been active on, I've seen both approaches. oss-security
and all the mplayer and ffmpeg lists are other examples that use(d)
Reply-to. Busybox and uclibc and libc-alpha (glibc) are some that
don't. I can use both (especially now that I found a good solution for
avoiding messing up replies myself) but I pretty strongly prefer the
use of Reply-to, because it tends to avoid having people accidentally
reply off-list and losing the continuity of threads on the list. And
since it's easy to detect Reply-to generated by the list (e.g. just
look for the To and Reply-to addresses matching), any good client
should be able to override this default for power users who really
want to override it.

Rich

Re: Mutt group reply

[Rob Landley]

On 07/15/14 10:12, Rich Felker wrote:
> On Tue, Jul 15, 2014 at 09:27:09AM -0500, Rob Landley wrote:
>> On 07/13/14 22:51, Rich Felker wrote:
>>> On Mon, Jul 14, 2014 at 12:58:59AM +0400, Solar Designer wrote:
>>>> On Sun, Jul 13, 2014 at 07:59:04PM +0100, Laurent Bercot wrote:
>>>>> On 13/07/2014 17:34, Solar Designer wrote:
>>>>>
>>>>>> An alternative is to reconfigure the list so that it doesn't set the
>>>>>> Reply-To header, but this may result in many replies being inadvertently
>>>>>> sent off-list.  I think it's better for Mutt users to adopt a habit to
>>>>>> answer that question with "n".
>>>
>>> Thanks! I had been looking for a solution to this issue for a long
>>> time but didn't bother to really look into it.
>>
>> Thunderbird's "reply all" will _only_ reply to the reply-to, and I wind
>> up manually copying in individual email addresses to cc: when I bother.
>> (Yes, it has a 'reply list' button, but the reply-to header overrides
>> the difference.)
> 
> Reply-to headers should not override the 'reply to all' feature in a
> mail client. If they do, this is a bug. What use is 'reply to all' if
> it behaves the same as plain 'reply'?

Agreed. It's a bug. My guess why reply-to can still trigger such an
obvious bug in a project this old and widely used would involve reply-to
being obscure and seldom used.

(Note: on _this_ email, reply all worked. Because I wash my email
through gmail for the spam filtering and download it via pop, and doing
so has a duplicate killer you can't disable, and this time the direct
copy rather than the list copy arrived first. Since the reply-to is
added by the list, a user replying to the copy they were cc'd on would
never see a reply-to header in the first place. Which is one reason
_why_ it's so seldom used; it doesn't actually work. If somebody isn't
subscribed the list and you cc: them, their reply won't have a reply-to
to force it to the list. So the people most likely to _need_ reply-to
don't get it...)

*shrug* Not a big deal. I mostly read the list archives on the web anyway...

Rob

Re: Mutt group reply

[Felix Fietkau]

On 2014-07-15 17:12, Rich Felker wrote:
> On Tue, Jul 15, 2014 at 09:27:09AM -0500, Rob Landley wrote:
>> On 07/13/14 22:51, Rich Felker wrote:
>> > On Mon, Jul 14, 2014 at 12:58:59AM +0400, Solar Designer wrote:
>> >> On Sun, Jul 13, 2014 at 07:59:04PM +0100, Laurent Bercot wrote:
>> >>> On 13/07/2014 17:34, Solar Designer wrote:
>> >>>
>> >>>> An alternative is to reconfigure the list so that it doesn't set the
>> >>>> Reply-To header, but this may result in many replies being inadvertently
>> >>>> sent off-list.  I think it's better for Mutt users to adopt a habit to
>> >>>> answer that question with "n".
>> > 
>> > Thanks! I had been looking for a solution to this issue for a long
>> > time but didn't bother to really look into it.
>> 
>> Thunderbird's "reply all" will _only_ reply to the reply-to, and I wind
>> up manually copying in individual email addresses to cc: when I bother.
>> (Yes, it has a 'reply list' button, but the reply-to header overrides
>> the difference.)
> 
> Reply-to headers should not override the 'reply to all' feature in a
> mail client. If they do, this is a bug. What use is 'reply to all' if
> it behaves the same as plain 'reply'?
Thunderbird isn't the only Mail client that's affected. As far as I
know, Apple Mail and GMail are affected in pretty much the same way.

>> So it breaks other mail clients too. Largely because reply-to seems to
>> be used so seldom, and thus isn't particularly debugged. (This is the
>> only list I've followed in the past 5 years at least that uses reply-to.)
> 
> In lists I've been active on, I've seen both approaches. oss-security
> and all the mplayer and ffmpeg lists are other examples that use(d)
> Reply-to. Busybox and uclibc and libc-alpha (glibc) are some that
> don't. I can use both (especially now that I found a good solution for
> avoiding messing up replies myself) but I pretty strongly prefer the
> use of Reply-to, because it tends to avoid having people accidentally
> reply off-list and losing the continuity of threads on the list. And
> since it's easy to detect Reply-to generated by the list (e.g. just
> look for the To and Reply-to addresses matching), any good client
> should be able to override this default for power users who really
> want to override it.
I consider lists using Reply-To to be badly broken. Is it really worth
breaking often used regular features (like either replying directly to
the author, or reply-all) with several popular mail clients, just for
the sake of preventing a few accidental off-list emails from people who
click the wrong button? I don't think so.

I think this is spot on: http://www.unicom.com/pw/reply-to-harmful.html

- Felix

Re: Mutt group reply

[Rich Felker]

On Thu, Jul 17, 2014 at 07:48:36AM +0200, Felix Fietkau wrote:
> > Reply-to headers should not override the 'reply to all' feature in a
> > mail client. If they do, this is a bug. What use is 'reply to all' if
> > it behaves the same as plain 'reply'?
> Thunderbird isn't the only Mail client that's affected. As far as I
> know, Apple Mail and GMail are affected in pretty much the same way.

GMail's mishandling of threading makes it pretty much completely
unusable for working with mailing lists (albeit more convenient for
most non-list email usage) so I don't think its behavior is all that
relevant. Not sure about Apple Mail.

> I consider lists using Reply-To to be badly broken. Is it really worth
> breaking often used regular features (like either replying directly to
> the author, or reply-all) with several popular mail clients, just for
> the sake of preventing a few accidental off-list emails from people who
> click the wrong button? I don't think so.
> 
> I think this is spot on: http://www.unicom.com/pw/reply-to-harmful.html

I'm quite aware that this is a controversial topic with a long
history, and I've read all the arguments plenty of times before. To
answer your specific questions, if the intent is to actively prevent
off-list replies except when someone goes out of their way to do one,
then the Reply-to header added by the list is simply doing its job.
That's why I like it. Certainly there are a few special situations
where it's appropriate to take a reply off-list, but they're the
exception not the norm. (And the result reminds me of what happens on
IRC when users start private queries with the first person who
indicates they might have knowledge on a topic rather than keeping the
question in the channel. :-)

Rich