Re: [tlug] reply-to headers

["Stephen J. Turnbull" <stephen@example.com>]

Curt Sampson writes:

 > Also, I do believe you've misinterpreted Stephen here. Your reply
 > makes sense in the context of an MUA processing headers, but not in
 > the sense of generating headers, which latter case is what I
 > believe Stephen was talking about.

That's right.  There is a reasonable algorithm for generating a
default set of addresses for use in a group reply:

1. If there is a reply-to header, reply to that, exactly.

2. Otherwise, if there is an RFC 2369 list-post header, reply to that
   and CC: all explicit recipients except self, one per line (in the
   UI).

3. Otherwise, reply to the author, and CC: all explicit recipients,
   one per line.

An author who wants an explicit reply in a list context should set
Reply-To himself.  If this algorithm would be widely used, the
tendency of CCs to accrete in long threads would actually be reversed
(because of the "except self" clause in #2), and authors would rarely
find themselves needing to trim CCs.

Note that this algorithm actually works well for ordinary
conversations as well.  So as long as you can distinguish between "I
want to reply privately" and "I don't care if this is public", it's
quite easy to use the "reply" function for truly private replies,
going only to the author, and "generic" replies, which GTTRP (Go To
The Right Places).

 > Basically, anybody who doesn't want two copies of a reply has the
 > choice of setting his reply-to header to send things to the list
 > only.

This isn't quite true.  You can also set Mailman's per-recipient
"not-me-too" flag, in which case if your subscribed address appears in
the addressees, Mailman will not send you a copy.  This has the defect
that you won't have access to some proposed improvements (specifically,
a header that tell you where the list post will be archived), but
otherwise works well for most people.

 > It appears that a lot of people who don't want two copies refuse to
 > configure their own MUA to do do this, and instead demand that the
 > list software be configured to set their personal preference for
 > everyone.

I don't really think that's the motivation.  Rather, people
(especially list policy setters) want to strongly encourage replies to
the list, as if it were a web forum.  This is the explicit statement
of the PTBs at TLUG, and the invariable claim of other wrong-headed
supporters of reply-to munging.  I see no reason to do anything but
take them at their word.

Users who want it are just used to using reply-to-author to reply to
lists; I think that if all lists stopped using it, they would learn to
use reply-to-all to list traffic very quickly.  They would even like
it a lot if the above algorithm were used.