[pmwiki-users] Mailling list Reply-To:

[Patrick R. Michaud]

Oliver Betz writes:
> is there a prevailing opinion in this list whether the list software
> should set the Reply-To: to the list address or leave it untouched?

This is an excellent question, and I was surprised and pleased
this morning to see so many useful responses to it in such a 
short period of time.  

Since the topic comes up from time to time on the pmwiki-users 
list (as well as other lists I work with), I figure it's a good
opportunity to write up my analysis, opinion, and responses on
the topic, and we can then point to this message or an article 
on pmwiki.org if/when it comes up again in the future.

For those who just want the bottom-line answer, I think that
things are better if we don't set the Reply-To: address for
the mailing list.  I don't have any qualms with mailing lists 
(or administrator) that do choose to set Reply-To:, but I don't
want to do that myself for my own lists.

With that, here's my comments to some of the responses in the
thread, and my reasons for deciding against setting Reply-To 
on pmwiki-users.

Nils Knappmeier wrote:
> I wouldn't like [Reply-To munging]. When I post something to the 
> list, and people reply to this post, then I'd like to have a 
> copy of this answer outside of my pmwiki-folder. 

I have to agree with Nils on this one -- I particularly like
that responses to messages I've written come more quickly and directly 
to my inbox via the 'Cc:' line.  There are times when the mailing
list server gets a little backed up in sending out messages, or when
it gets hung up on delivering to someone else's address before getting
to mine.  Several times I've had the situation occur where someone
responds on the list to a message I've written, but delays in mail 
processing mean that I start seeing others' replies to that response 
long before I ever get the original response that they're replying to!  
If my address is in the Cc: line, I get replies directly (and
immediately) for any threads I've contributed to.

Kathryn Andersen writes:
> Get a decent mail program that has a reply-to-list option 
> (I use mutt, myself).

Robin and Joachim Durcholz both responded with:
> Get a decent mail program that has a reply-to-author option

I'm not trying to pick on Kathryn, Robin, or Jo, but I see
these a lot in the discussions on Reply-To, both here and
elsewhere.  Frankly, I find it more distracting than helpful -- 
along the same lines as someone saying "switch to 
Firefox/Internet Explorer/Mozilla/Opera...".

I have two basic observations to make on this point:  
(1)  I totally agree with the part of mutt's tagline that 
says "All mail clients suck", and (2) people choose 
(or are constrained to use) specific mail clients for 
reasons that are often far more compelling than the 
set of "reply" options that happen to be available in 
the client.

I also want to note that "reply to author" in many clients
can't really work correctly (i.e., according to the standard) 
if the original author had a Reply-To: header in his message 
but the mailing list munged it out of existence.  Although I 
do grant that this isn't a common occurrence.


Ultimately I don't think that mailing lists should set Reply-To
headers because the RFC 2822 standard is pretty clear
that Reply-To "indicates the mailbox(es) to which the author
of the message suggests that replies be sent." [2]  The mailing
list isn't the author of the message, so it shouldn't be 
meddling with the Reply-To header.

More to the point, there *is* a header that mailing lists
should be using to direct replies -- the List-Post header
from RFC 2369 [3].  What's more, all modern mailing list
managers are already using this header, including the
one running the pmwiki-users mailing list.

So, in order to get the list to match user expectations 
(i.e., "replies to messages coming from the list should go 
to the list"), what mail clients should be doing is to set 
the common/generic "reply" function to give priority to any 
address given in the List-Post header if it exists.  
Of course, a mail client should _also_ provide separate 
"reply-to-author", "reply-to-list", and "reply-to-all"
options, as well as give the user the ability to customize
the meaning of "reply", but for some reason this all
seems to be asking too much of mail client software.

The principle is simple -- it's the *mail client* that should
make the users' "reply" buttons conform to what they expect
it to do, because the mail client sits closest to the user.  
Mailing lists should not be second-guessing user preferences;
they should conform to the published specs for mail, just as 
we expect web servers and web pages to conform to the published 
specs for HTTP and HTML.

Now then, in saying the above I do recognize that sometimes reality
dictates that we deviate from standards, such as when there
is a large population of broken browsers ("Internet Explorer")
or broken mail clients ("all of them [except the one I'm using]").
In such cases following the standard can abandon or alienate
large portions of the intended audience, and it's why I don't
have any qualms with websites that use "IE hacks", tables for
layout, or mailing lists that set Reply-To.  The people going
outside of the standard are just recognizing reality as they see it.

But in the case of the pmwiki-users mailing list, following the 
standard doesn't appear to be a major impediment to people's
ability to use the list, and several have expressed that not 
following the standard would be an inconvenience to them
(and I'm among the "would be inconvenienced" group).  So, in 
this case I think it's better to have pmwiki-users follow the 
published standards than to perpetuate practices that go 
against the standards.  

Hopefully someday we'll have mail clients that agree, but I'm
not holding my breath.

Pm

P.S.:  Some classic articles in this area are:
    Rosenthal, C., "Reply-To Munging Considered Harmful" [4]
    Hill, S., "Reply-To Munging Considered Useful" [5]
    Pickette, N., "Reply-To Munging Still Considered Harmful. Really." [6]

 
References: 
1.  http://www.mutt.org
2.  http://www.ietf.org/rfc/rfc2822.txt
3.  http://www.ietf.org/rfc/rfc2369.txt
4.  http://www.unicom.com/pw/reply-to-harmful.html
5.  http://www.metasystema.net/essays/reply-to.mhtml
6.  http://woozle.org/~neale/papers/reply-to-still-harmful.html