<<< Date Index >>>     <<< Thread Index >>>

Re: mutt/2304: reply / group reply behavior broken WRT $reply_to and $reply_self



The following reply was made to PR mutt/2304; it has been noted by GNATS.

From: Derek Martin <code@xxxxxxxxxxxxxx>
To: bug-any@xxxxxxxxxxxxx
Cc: 
Subject: Re: mutt/2304: reply / group reply behavior broken WRT $reply_to and 
$reply_self
Date: Tue, 27 Jun 2006 07:05:00 -0400

 --at6+YcpfzWZg/htY
 Content-Type: text/plain; charset=iso-8859-1
 Content-Disposition: inline
 
 On Tue, Jun 27, 2006 at 08:05:02AM +0200, Thomas Roessler wrote:
 >  On 2006-06-24 18:55:02 +0200, Derek Martin wrote:
 >  
 >  >  No it isn't... at least not necessarily.  It is only the
 >  >  address where you want replies to go.  In addition to the
 >  >  other examples I gave, there is the very common example of
 >  >  mailing lists.  People who use mailing lists often set the
 >  >  Reply-to header to the mailing list they are reading.  If
 >  >  they reply to their own message, clearly they want the
 >  >  reply to go to the list; not to the recipient list.  
 >  
 >  Is that so clear?
 
 If it were not the case, again, would they not unset $reply_to?
 Assuming they normally have $reply_to set, we must also assume they
 normally want to honor the sender's Reply-to header.  I don't see any
 reason to assume the case is different if they are the sender...
 Again, with the current code, if they do want this behavior, expecting
 that setting $reply-to will in fact honor the Reply-to header (as the
 documentation clearly states that it will) they simply can't have it.
 
 >  The main assumption when $reply_self is unset is that people
 >  won't want to talk to themselves.
 
 But in making this statement, you're again assuming that they are
 setting Reply-to to some address which represents themselves, and I'm
 trying to get through to you the simple fact that this may not be the
 case at all...
 
 >  Also, Reply-To is just about the last thing you'd really want
 >  to use to redirect follow-ups to a mailing list.  But that's a
 >  different discussion.
 
 I happen to agree with this in general; but that doesn't change the
 fact that it is common practice, and there are also perfectly valid
 reasons to do it.  I've already given several.  Here's perhaps a
 clearer scenario (if you already get the usage cases, feel free to
 skip down below regarding the actual behaviors):
 
 Someone at a company is replying to a private e-mail, and may want the
 private discussion he is having re-directed to a mailing list, because
 it is a more appropriate forum for the discussion.  So he sets the
 Reply-to header to the mailing list.  It might not even be a mailing
 list; it could simply be a company's in-house alias that points at all
 the right support people for the other party's request/question.  The
 original sender may have no knowlege of the list, and no desire to be
 a permanent member of the list.  Using Reply-to is a good way to get
 that mail going where it needs to go...
 
 Now, in such cases, if the person who set his reply-to header wants to
 continue to participate in the discussion, because he has an
 afterthought which is relevant, he will want all follow-up mails to
 also to to that same reply-to address, so that everyone else knows
 what he is discussing.  But what are the odds he will want to continue
 to reply to the same thread (to his own message), and have only
 himself or only the original sender as the recipient?  It's possible,
 though I think the percentage is relatively low.  
 
 With the current code, the more common case is not possible.  Whether
 $reply_to is set or not, the behavior of Mutt depends only on
 $reply_self.  If $reply_self is set and the user uses Reply, mutt will
 send the reply only to the user.  If it is unset, mutt will send the
 reply to the original recipients.  But if $reply_to is set, this
 behavior does not change!  How does that make sense?  You already get
 the behavior you are saying you want, if reply_to is unset.  But it is
 not possible to make mutt honor the Reply-to.
 
 >  Staring at the current code, I don't see the behavior that you
 >  describe -- but maybe that's because I'm sitting in a hotel and
 >  waiting for breakfast to be served.
 
 I can't follow your analysis; I find it rather difficult to follow
 code which is broken up in an e-mail that way.  I can tell you that
 your conclusion seems to be wrong...  I suggest rather than trying to
 analyze the code, you actually send yourself a test message and try
 replying to it with the two variables set and not set, and make a
 table of the behavior.  You'll find that for three cases, the behavior
 of mutt does not change regardless of the setting of $reply_to.  What
 you need to do to test is this:
 
   1.  Send yourself a message.  
   2.  Include two addresses on the CC list which are not you (remove
       from alternates if needed, or send them to me if you want).
   3.  Include a reply-to header which goes somewhere convenient for
       testing, which is NOT a mailing list you are subscribed to, and
       not your own address.  It doesn't matter what it is, since you
       don't actually need to send any mail there...
 
 Reply to this message, with $reply_to unset.  Note the behavior with
 $reply_self set, and then unset.  Now set $reply_to and repeat.  You
 will see that the behavior does not change; the two recipient lists
 will be the same, regardless of $reply_to.  I can't see how this is
 not a bug.
 
 Repeat again, using Group-reply.  This time you will notice that the
 behavior DOES change in one case, but not in the other.  Still a bug,
 and even worse because it is inconsistent...
 
 My patch fixes all three cases.
 
 -- 
 Derek D. Martin
 http://www.pizzashack.org/
 GPG Key ID: 0x81CFE75D
 
 
 --at6+YcpfzWZg/htY
 Content-Type: application/pgp-signature
 Content-Disposition: inline
 
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1.2.1 (GNU/Linux)
 
 iD8DBQFEoRDcHEnASN++rQIRAgyKAJ4pKGMI/POesxdfis6eVNFNpCJgSACgtvPt
 s/OthtdkwJzBqT3ucqFVBCQ=
 =XcCM
 -----END PGP SIGNATURE-----
 
 --at6+YcpfzWZg/htY--