I'll admit right out that I'm "new" to email (meaning that I've only been using it for ten or twelve years, never implemented an MUA or an MTA, and don't administrate any mail servers), but there's this meme titled "Reply-To" Munging Considered Harmful (started by one Chip Rosenthal) that I would like to take issue with. The meme has managed to worm its way into mailing list administrators worldwide, and I even thought I would be one of them, waving the standard of "well-behaved Internet software" as I marched on. But it has not come to pass.
There are two mailing lists I'm on. One, the pygame-users mailing list, munges the Reply-To header. Another, the pyode-user mailing list, does not. On every mail I have replied to on the pyode mailing list, I have sent a follow-up to the list saying "oops, sorry Person-I-Replied-To for the duplicate mail, this was supposed to go to the list too". This hasn't happened once on the pygame mailing list. Mr. Rosenthal asserts that reply-to munging "adds nothing". This is not true. In my case, it adds relief from embarassing "shit I fucked up with the email thing" notes and copies sent to lists.
In most computer-mediated fora, if someone says something, and you'd like to respond, it is expected that the response will go to the same place that the original did (meaning: sent to the same people). Mr. Rosenthal asserts that any reasonable mailer has two functions, one for "reply" and one for "group reply", and that any idiot user can be taught to press "g" if they want to respond publically, and "r" if they want to respond privately. But a user expects that a "reply" will be sent to everyone relevant; it is the private responses that are special, not the public ones.
Back in the '90s, you frequently found that some AOL user had sent you and 400 other people a message, and one of those 400 other people had responded with "I don't want to read this crap", and another of those 400 other people had written back with "lol", etc., with each message growing longer and more meaningless until you gave up and got a different email account. This was entirely the result of making "reply-all" the default function. Idiot users who didn't know better would send replies to people who didn't need them. So for the most part I think making "reply privately" the default function for end-user software is a defensible choice; it prevents unwanted noise. But when you start a mailing list, you want to make the default reply function (on a per-message basis) "send to list"; it is for this purpose that reply-to munging is helpful.
This is the case for reply-to munging. Let's go through Mr. Rosenthal's arguments against.
It breaks the reply-privately function. It's not gone, only more difficult. This isn't always a make-or-break thing; if nobody ever replies privately, then it's a tradeoff that should be considered.
Administrators shouldn't dictate that all replies go to the list. This is a straw man. Administrators know best what the tone and flavor of their mailing list is. If it's primarily technical discussion, or a digital gathering space that doesn't usually wander off into private email discussions, then absolutely they should have the power to make replies go to the list by default.
Reply-to munging can make it difficult or impossible to find the sender of a message. As someone who is new to email, I'm not familiar with the uses Mr. Rosenthal offers: if someone is sending a message that they haven't authored, typically they indicate this fact in the body of the email and include the content of the message as a quote. If someone is sending email from one address, but would like to recieve email at a different address, they should change the From: header to match the Reply-To: header. (I know it's really more complicated than that, but I think this approach is viable.)
Munging reply-to headers penalizes those who use good software. Mr. Rosenthal writes, "If a few people need to type in a full reply address so that everybody else can use all the features of their mailer, I say, 'Fine!'" To this I reply: if a few people cannot use all the features of their mailer so that everybody's replies go sensible places by default, I say "Fine!"
Munging reply-to headers makes more work if you want to reply privately. On the pyode mailing list, hitting "Reply" will reply To: the person who sent the mail; hitting "Reply all" will go To: the sender and Cc: the list. Most users (including technically inclined ones) will assume this means that the sender will get a duplicate email from the list, and so will scratch out the To: line and change the Cc: to a To:. So not munging reply-to headers leads to users doing just as much work, assuming they even remember to reply publically. (Otherwise they have to go to their copy of the sent message and send a copy to the mailing list.)
In fact, Mailman is intelligent enough to not send a duplicate email in this case, but frankly this is a huge surprise to me considering that Mailman sends passwords in plaintext emails. However, if B replies to A via the list L, the reply will be To: A and Cc: L. If C replies to this email, it will be To: B and Cc: A and Cc: L. This is just ridiculous.
A user expects a garden-variety reply to be private. This is just not true.
A user may expect a reply to be sent privately, only to discover it going out as public, which can lead to embarassment or worse. I have never seen this happen, and it has never happened to me.
If we munge reply-to headers, users will demand a third "reply" option which ignores reply-to headers but addresses an email to the sender of a message. This is complicated, but: if you already accept that there is a category of "private" replies and a category of "public" replies, why shouldn't the "private" reply in this case write to the sender? Anyhow, there's already a third type of "reply" called Reply-To-List, and it doesn't work, so I don't see what good Mr. Rosenthal's suggestion is.
In an ideal world, people would never make mistakes, and they would know that their mailer has reply-private and reply-public modes, and would invoke the correct ones all the time. That isn't the world we live in. In a slightly more plausible world, mailers would have a default "reply" mode which intelligently decides who gets a copy, based on who got the original message, whether it's part of a mailing list, and how dumb the user is, and explicit "private" and "public" modes in case the software's guess is wrong. We don't live in that world either. We also don't live in a world where users expect replies to be private by default. Munging reply-to can be a terrible thing, or it can be benign or beneficial. List administrators should not blindly follow Mr. Rosenthal's suggestion, but think about their users and what is likely to be best. For the lists I administrate, I'm enabling munging.
Lastly, email has grown to be a hugely complicated organism, made up of many users, many pieces of mail software, and many standards. We can't throw the whole thing out and start fresh; even if we did, how would replies work, given that many users are prone to noise generation? Reply-to munging is just a patch, but I think it's the best solution we have, and I'm going to use it when it's called for.