In the past two weeks I have been discussing the use of Reply-To headers on mailing lists on several occasions. Some people are afraid that people are not be able to use their MUA (mail client) in a proper way and insist that mailing list software administrators configure that lists in such a way that the software adds a Reply-To header so that replies automatically end up on the list. They argue that this is important in order to keep discussions on the mailing list. Those same people also often argue that it is a good thing to do because it prevents duplicates being received when somebody replies to a message that was sent to the list. There are a few major flaws in this argumentation which I would like to illustrate.

Mail clients have often two ways of replying to e-mail messages: reply-to-sender and reply-all. In default cases (case 1), if a person sends a mail to a mailing list all participants see a mail with a From header from the user, and as To header the mailing list. In case Reply-To headers are added (case 2) the participants see a From header with the sender's email address, a Reply-To header containing the mailing list's address and a To header containing the mailing list's address.

If in case 1 a message is to-be-replied to the original sender, a user only has to hit "reply" . In almost every MUA "reply" defaults to "reply-to-sender". To reply to the whole mailing list the user has to use "reply-all". In that case the original sender might receive a duplicate mail message (once as normal recipient through the To header, and once through the mailing list). Luckily all modern mailing list managers (like mailman) allow you to configure a feature that excepts users from receiving a mail through the mailing list if there address is found in a To or Cc header. This "filter-out-duplicates" feature conveniently breaks down the argument of the duplicate received messages that Reply-To-header-adding proponents bring up.

In case 2 the user just have to use "reply" to send his reply to the whole mailing list, "reply-all" will do the same. This clearly shows that using a reply-to header makes things easier for users. There will also be no duplicate as the original sender's address won't appear in the headers of the message that is replied. However, some mail clients (like pine and mutt) do only have "reply" (which defaults to "reply-to-all"). For those clients there is now suddenly one extra question whether the reply should use the address in the Reply-To header, or the address from the original From header. This is somewhat of a nuisance that can be fixed by a simple procmail filter in combination with the formail utility that removes the Reply-To header:

:0 f:formail4.dummy
* ^List-Id: <phpdbabstraction
| formail -I "Reply-To"
:0:
* ^List-Id: <phpdbabstraction
mail/dbabstraction

There is a technical argument against munging headers, which is written down in some RFC (which I can't find right now unfortunately). This RFC states that headers should only be added to mail messages in case it is of vital importance to the delivery and routing of the mail message.

Adding a Reply-To header will overwrite an already existing Reply-To header, meaning that that the sender might not actually the reply on the address he/she wanted a reply to be send back to. Some people are unfortunately misguided and don't care so much about this technical concept.

However there is a sociological problem with Reply-To header munging as well, which I would like to illustrate with some examples. Just like other somewhat larger companies do we have an internal mailing list where we as employees can discuss issues of various matter (both "useful" and "fun" things). Subjects that can be found on this list are for example subjects that concern everybody in the company, such as new marketing material, but also invitations to film showings, "interesting" news items, etcetera.

In some cases not everybody is fully agreeing with some of the ideas or questions that are shared on this list, and this can result in somewhat heated discussions­. During heated discussions people pay by nature less attention on what they are sending with their mail client and forget to reply to the original sender only (by editing the headers in the reply message) and instead reply to the list with words that are not thought out well enough to be shared with everybody on the mailing list. A worse example that I encountered myself is a misdirected mail calling some of the people on the list (a business partner in this case) ass holes.

There is no doubt that this can result in quite a bit of damage for the person that was supposed to send the reply only to the original sender. For others this might be a form of entertainment because besides the "personal attack" replies there are also examples of more humorous nature where a reply message ends up on the whole mailing list. For example a reply with a question if the intended person of the reply will go out on a date, or a reply that almost suggest there is some lesbian relation going on.

To prevent the kind of madness that gets people into problems, the principle of least damage should be applied. There can never be a problem if a mail is reply by accident only to a single person and not the whole mailing list, because it is trivial to resend the message to the whole mailing list. (It is still an art to do that properly though as people forget about "threading" - more about that in a later post). However, it is always a problem if a personal reply ends up on the whole mailing list with text that should not have been disclosed to all mailing list recipients. Therefore I would urge people to not to use Reply-To munging on their mailing lists--even if this decreases the entertainment value :).

There is also some more information here . (This resource contains more information about older mail clients and Usenet usage, the arguments being made still apply for normal mailing lists as well).