Recently I posted a comment on Stack Overflow on an answer by (say) Felix. He replied to the comment, and when replying to him I started my comment with "@Felix". Stack Overflow stripped that part of my comment out.

The engine should leave that in place, even if I'm redundantly specifying the person attached to the answer. Sometimes, regrettably and hopefully temporarily, discussion ensues and so it's useful to see to whom a comment is directed. It's inappropriate for the engine to remove an explicitly-typed direction in this way. Without wanting to blow one's own horn, the engine isn't smarter (in most cases) than the person typing the comment, and should leave it alone, complete with the "@xyz" part at the outset if explicitly typed.

Note that this is not a duplicate of "Eeeeek - what happened to my @ salutation?". That is asking why. I'm not asking why. I'm saying it's wrong and should be fixed.

See also: Tell the user who will be notified of a comment

Update 2011/07/21: Having made my request and reasons clear, I had stepped back from this and just lurked, since I didn't really have anything more to add. But others did, and I got re-engaged, etc., etc. Since my original edit and update were a bit rambling, I'm replacing it with what I believe to be the main points in favor of and against keeping the change.

Separately, I realized that I'd failed to do the obvious thing and suggest an alternative, and so I've done that now: Tell the user who will be notified of a comment.

Apologies in advance if anything has been lost in translation/condensation (feel free to fix it, just try not to be even more verbose than I am):

Arguments in favor of keeping the change:

  1. @postowner is noisy. Protecting the signal-to-noise ratio is vital to keeping the standard high on the Stack Exchange network. (more) If you want to specifically address the postowner, just use postowner without the @, e.g. joe: or Joe,.

  2. @postowner teaches new users that they have to do that to notify the postowner, and of course, they don't. Just commenting on a post always notifies the postowner.

Arguments against keeping the change itself:

  1. @postowner is not noisy. postowner is no less noisy than @postowner. This does nothing to improve the signal-to-noise ratio.

  2. @postowner is not just to notify the postowner, we use @xyz a lot like names on Stack Exchange sites and about 18 million other places online. The @ convention didn't originate at Stack Overflow and isn't unique to the Stack Exchange network. To ignore this is to ignore the human factor.

  3. Forcing their removal reduces clarity. Sometimes when you comment on a post, you're commenting to all readers; other times, you're commenting specifically to the post owner and so the natural thing is @postowner as with everywhere else we direct comments (on Stack Exchange and elsewhere). Their removal breaks the distinction, forcing workarounds.

  4. Workarounds for the change cause more harm than making the change prevents. If users see a lot of postowner: or @ postowner (note the space) or postowner, they mistakenly learn the @ is generally unnecessary and leave it out when they do need it, preventing the person they're trying to direct the comment to from being notified. This is active harm.

  5. It's inappropriate for automated systems to silently edit content from actual intelligent beings, especially without a means of overriding that edit, without an overwhemlingly good reason, which this is far from being.

  6. It's too intrusive, feels like censorship and editing for style, turning off new users who might otherwise contribute. One person's "noise" is another person's essential social lubricant.

  7. It makes it harder to teach by example / is confusing for new users / is over-complicated (even for established users). (more) Sometimes you need the @, sometimes you're actively prevented from using it. Quoting Jeff:

    When the answer to a problem is "let's make things arbitrarily more complicated", it is rarely the correct answer

  8. It forces asymmetry if there is a degree of back-and-forth. (more & example)

  9. Finally: In the face of overwhelming, unambiguous negative feedback, it would be simply inappropriate to keep the change even if the 90% expressing their disagreement with the change were wrong. (more) Its benefits are nowhere near sufficient to justify overruling that kind of majority, especially not given that "We don’t run Stack Overflow. The community does." Overriding the community on this is active harm.

  • 35
    #T.J. I say we should just use # for addressing the poster Jul 10, 2011 at 2:04
  • 1
    Well done for trying and picking up the glove but as was the case with the envelope R.I.P (where many members wanted it restored) changing decision of the Developers once such a decision was made is impossible. Jul 10, 2011 at 10:54
  • @Shadow, there have been instances in the past of declined posts turning into completed posts.
    – Pops
    Jul 13, 2011 at 19:23
  • @Popular I doubt this one will become one of these.. the cases I saw are new ideas here we're talking about restoring previous behavior. Jul 13, 2011 at 20:05
  • @Shadow, yeah, I'm not hopeful for this one either, but I thought it was worth pointing out since you used the word "impossible."
    – Pops
    Jul 13, 2011 at 20:55
  • @Shadow, @Popular: The envelope thing was a botched enhancement and total UX failure that was then corrected. Correcting it makes up for everything. The system we have now is a good one; for a short time, we had something that was screwed up, but it didn't last. Jul 13, 2011 at 21:15
  • Good point @T.J. but still my points stands and the problem is that Jeff consider every non vital word in question, answer or comment as "noise" and we can't really change the way he thinks. Jul 13, 2011 at 21:27
  • 12
    @Shadow: True. But while in the early days what Jeff thought should almost certainly have been law -- that's how you establish something, with a clear vision and strong guidance -- three years later, if the claim is that "the community runs it," it's time to let the community do exactly that. This kind of micro-management can very well kill the SE network. It's grown beyond one man's vision, and it needs to. And again, if you're going to flaunt the community, pick something worthwhile. One @ is not remotely worthwhile. Jul 13, 2011 at 21:32
  • 2
    Thanks for bounty. I wonder if there's a way to re-open this now that Jeff is long gone? Jun 20, 2015 at 4:57
  • 1
    And over 4 years after the implementation, it seems even SE developers are still confused. Like “Comments should be at least 15 characters in length” does not make clear @name might be excluded shows that a developer made the Android app automatically add the @name part when replying to a comment, after which the same thing is silently ignored when counting the comment length.
    – Arjan
    Sep 6, 2015 at 12:56
  • 1
    What I love about this is how many comments here start with an @username. LOL.
    – Millard
    Jun 24, 2021 at 20:12

12 Answers 12


This change also makes it harder to teach by example. When a new user enters a site, I make sure to reply to them in the form "@newuser: What do you mean by X? Y or Z? (Use @Jan in your reply and I get a notification)" This teaches them right from the start how the commenting system works. I know that you also get a notification without an @lert when no one else joined the conversation (yet), but this is an exception to the rule, for all those new users that don't know the system yet. We should not make the exception more prominent.

"@postowner, you forgot to provide some information" is also a pretty natural sentence, so it's not that noisy. I would not object to removing it from the global inbox notifications, but please leave it in them comments themselves.

  • 8
    the help link available under every [Add Comment] button already explains the basics of replies, without littering comments with thousands of noise words. Also FYI, we detect when only two people are talking and notifications will work regardless. Try it... Jul 3, 2011 at 9:38
  • 4
    furthermore, what you describe actually "teaches" new users something incorrect -- that is, the post owner is always notified of every comment, and seeing other "experienced" users choose to type @postowner would imply otherwise. Jul 5, 2011 at 7:49
  • 13
    @Jeff: notification and content are two completely different things. Please don't try to argue that one implies and restricts the other.
    – Jason S
    Jul 8, 2011 at 16:35
  • 29
    @Jeff, you're applying a technical solution to a social problem. At some point the newbie will need to learn the @ protocol, and by introducing inconsistencies into the system you've made that much harder. As for teaching something incorrect - well it's only incorrect because you declared it so. You could easily change your mind. Jul 8, 2011 at 21:40
  • 1
    @mark it isn't inconsistent; the post owner is always notified of every comment on their post, forever, always, in every possible circumstance, no matter what the contents of that comment are. The optional part is notifying someone else which is what @username is about. Jul 8, 2011 at 22:11
  • 45
    @Jeff, it is inconsistent in that sometimes you need it, sometimes you don't and sometimes it's deleted, sometimes it isn't. It doesn't matter that there's a logical explanation, it's still an inconsistency in the interface. Jul 8, 2011 at 22:19
  • 1
    @mark it is not inconsistent; you never need @postowner to notify the owner of a post about your comment. Period. Jul 8, 2011 at 22:21
  • 22
    @Jeff, what I mean is that it's inconsistent depending on whether you're commenting to the post owner or to someone else. It's a UI inconsistency, not a logical inconsistency - I get your point too. Jul 8, 2011 at 22:35
  • 35
    @Jeff: Who cares how the notification system handles this edge case. I don't think any of us who object to the behavior has a strong opinion. We're talking about the written content of a comment. Don't mess with my content!
    – Jason S
    Jul 9, 2011 at 1:22
  • 20
    @Jeff, you can't rely on the help link because nobody reads anything. Now where did I learn that? Oh yeah, here: codinghorror.com/blog/2009/10/treating-user-myopia.html Learning by example is much more reliable. Jul 9, 2011 at 4:03
  • 15
    @Jeff, which is exactly the wrong thing to learn when they try to respond to someone else's comment a few minutes later. This is the essence of the inconsistency argument I'm trying to make above, and the whole point of the answer to which we're commenting. Jul 9, 2011 at 4:52
  • 3
    @Jeff, I'd like to continue this as a chat: chat.meta.stackoverflow.com/rooms/399/… Be kind to me, this is my first exposure to the Chat rooms. Jul 9, 2011 at 5:07
  • 6
    I find it really ironic that all of these comments are prefixed with "@username". Isn't that proof enough that this change should be revoked? Aug 6, 2011 at 1:33
  • 6
    @Jeff The prefix is there for context and clarity. Do you go about deleting parts of answers just because they take up space? Signatures are one thing; they don't add any value to the question. These "in reply to" prefixes do add value, because it's extremely difficult to follow comment threads of 4+ comments without them. I know the content here is cc-wiki, but please don't insult our intelligence by assuming that part of our contributions are inherently worthless and simply "take up space" Aug 30, 2011 at 6:38
  • 6
    @Jeff in addition, there are over 600,000 posts on SO that have at least four comments (so about 1 in 10), and nearly 50,000 posts that have at least 10. So even though the majority of posts won't benefit from the added context, the long tail of posts will. Isn't that what Stack Exchange is about? Catering to the long tail? ;) Aug 30, 2011 at 6:43

I agree.

  • The system should minimize instances where it automatically modifies user input.
  • If you feel that the system should automatically modify user input, it should allow the user to override. Not allowing the user to override is offensive. Let the user say what they want to say, and if it's a problem, address it by moderation.
  • If there is no mechanism to allow the user to override the system, at the very least will you please make it provide visible feedback (1) that it is doing so intentionally, and (2) why it is doing so. Otherwise you not only have a system that is not only inflexible and offensive, but also extremely confusing.

...and if the sysadmins maintain this as status-declined, then what I'll do is stop using "@" signs, because it just causes annoyance, and I don't want the system automatically censoring my comments.

And just for kicks, let's compare these two possibilities.

Lorem ipsum, etc.


That doesn't look right. --Barman

@Barman: Are you sure? --Fooman

@Fooman: Yes, I am. See "Book of Lorem Ipsum" p. 302 --Barman

@Barman: OK, if you insist. --Fooman

or this:

Lorem ipsum, etc.


That doesn't look right. --Barman

@Barman: Are you sure? --Fooman

Yes, I am. See "Book of Lorem Ipsum" p. 302 --Barman

@Barman: OK, if you insist. --Fooman

The first of these is symmetric, the second is not. It is really strange, when someone writes a comment to you with your @username, to be completely unable to respond in kind, without any explanation by the system whatsoever why your use of @username is removed but the other guy's is not.

If some of you think this is excessively noisy, then let it be a client-side preference to remove the @name syntax. But don't change the source comment. If I want to make it clear who I'm talking to, let me do it.

  • 4
    indicating that you are talking to the post owner, in the absence of any other people talking, is just as redundant as a "hi fellow coders!" salutation. All comments, by definition, notify the post owner in every possible case. Polluting signal with a bunch of redundant @postowner is the same as polluting signal with a bunch of "hi fellow coders!" Also, if you want to use someone's name without the @ symbol, nobody is preventing you from doing so. Jul 9, 2011 at 4:59
  • 51
    @Jeff: You've said about a dozen times here that "@postowner is the same as...'hi fellow coders'". I think we all understand that you think that. What we're saying is that we disagree. Really quite a lot of us, including people who contribute at a high level to the communities. Regarding your last point, I had to respond to that in the question itself. Jul 9, 2011 at 8:57
  • @t.j you can read here for a lot more detail: chat.meta.stackoverflow.com/rooms/399/… Jul 9, 2011 at 9:07
  • 34
    @Jeff: Oh, great, let's try to make this even more disjointed and fragmented. "I'm simply right about this." I'm sorry, I really am, but you are simply wrong about this. "@Joe" is noise but "Joe" is not? No, that's absurd, it's not even a close call. And I haven't seen any links to "specific harm" in this disjointed conversation. I have seen links to discussions of too much discussion being harmful, which is yet another straw man, I'm not disagreeing with that. Jul 9, 2011 at 9:31
  • 5
    @Jeff: As with that parenting discussion, clearly your answer here isn't doing the job. I recommend another one, one which doesn't just constantly re-assert your claim that you're right, but actually lays out your points logically and completely, with those citations you talked about, including how you can possibly begin to justify the "@Joe" vs. "Joe" thing, how you respond to the complexity argument. In each case, with real arguments. Not straw men, not "I know I'm right." I also recommend re-opening your mind to the smart people here. Jul 9, 2011 at 9:34
  • @t.j I outlined my rationale at length in the above link. This is something I feel extremely strongly about, so we will have to agree to disagee on this indefinitely. Jul 9, 2011 at 9:36
  • 23
    @Jeff: If that link is it, it doesn't begin to do the job. Re agreeing to disagree: Hey, it's your site. Just remove the "We don’t run Stack Overflow. The community does." from stackoverflow.com/faq#reputation and similar if you're going to ignore what dozens of smart people are telling you. Jul 9, 2011 at 9:38
  • 36
    @Jeff I don't believe people enter the @ so people will be notified. I think people enter the @ because that is a convention that has evolved over years of BB and twitter use to indicate who you are talking to. That it will notify them is the last thing on my mind. I read the whole chat and don't think you have made a satisfying case to justify that @xxx is noise, and xxx is not. Nor that people enter the @ with the expectation that the person will be notified
    – JockM
    Jul 9, 2011 at 14:46
  • 8
    @JockM: "I don't believe people enter the @ so people will be notified" -- EXACTLY! I don't give a damn about how it affects notification. IMHO notification should be done when there are new posts on a question you posed, or new comments on an answer you posted or commented on, or on a question or answer that you explicitly decide you want to be notified about. To notify based on the @tags seems really risky. For one, usernames aren't unique, and for another, people don't always respond with verbatim @username tags (see @Jeff and @t.j in this thread).
    – Jason S
    Jul 9, 2011 at 17:16
  • 1
    @t.j. the specific harm is this: the average number of comments on a post is two. Multiply that by the ~8 million posts on Stack Overflow times the number of redundant @postowner words that are now present on our network. That's a lot of pointless noise. Also note that in the latter example Jason used above, no @usernames were required at all to get notifications as we detect when only two people are talking and auto-reply. Bottom line, unless someone explicitly indicates otherwise, comments are always directed at the post owner. That's the intent, and has always been from day one. Jul 21, 2011 at 10:47
  • 13
    @Jeff: Yes, you're clear that you believe that notifications are the only purpose of @postowner. There's a large number of counters to that on this page. But you continue not to address several points. 1. Even if you were right, it's wrong to push such a silly little thing in the face of this feedback. 2. That's not the only purpose of them. 3. The change is pointless, people just put @ postowner and postowner:, both of which are just as noisy, but which actively mislead new users for when they do need to use @ to direct their comment. I'm sure there are more. #1 is esp. big. Jul 21, 2011 at 11:03
  • 3
    @Jeff: I get your point, but it's nearly 100% confusing to anyone who hasn't read this webpage carefully. What you mean is essentially the following: that most of the time @user: syntax is appropriate, except in this one case of answerer + one commenter, where the commenter can't add the "@user:" syntax but the answerer can. How is a new user, who may be familiar with Twitter and Facebook and other social media where @user is the norm for addressing someone else, regardless of whether it has any automated effect, supposed to understand these rules?
    – Jason S
    Jul 21, 2011 at 13:27
  • 9
    @$#%@$#%@$%#@%!!!!! I just got a popup saying "Only one additional @user can be notified; the post owner will always be notified" thus denying me the right to use the literal @user syntax more than once in the comment even though I was using it in its literal and non-automatic form. You guys REALLY need to take your best shot at a simple, easy-to-understand webpage, and if you really feel the need to deny people posting a particular syntax, include a link to that webpage on the popup box so someone can actually understand what it is they're not supposed to do.
    – Jason S
    Jul 21, 2011 at 13:29
  • 2
    @Jeff: I had composed yet another reply, and then I got to thinking that things were really very, very fractured at this point and decided to put together pros and cons in handy lists. And somewhere along the way of doing that, the fact that I'd never offered a constructive alternative hit me upside the head. That's not how I like to do things. And I've come up with what I think (but then I would) is a pretty good idea: meta.stackexchange.com/questions/99261/… Jul 21, 2011 at 13:55
  • 7
    Quite ironic that Jeff uses the "@t.j" to start his comments...
    – chown
    Jun 25, 2012 at 18:53

Why must the classification of signal or noise be binary? Redundant addressing might legitimately fall between the two. Even if you do consider it noise, it is the most easily ignored noise possible, always in the same place with the same format.

I believe the downsides of this feature outweigh the upsides.

Edit: I believe the proper categorization of the @ tag is not pure signal or noise, but redundant signal. A redundant signal may be removed without changing the meaning of the message, by definition. However you will find that human speech is full of redundancy, because it enhances communication and improves understanding. I would argue for its retention in this case.


(This is coming at it from a different perspective, so I'm posting an answer rather than editing the question yet again.)

I think most if not all of the SE sites have the same statement Stack Overflow does: "We don’t run Stack Overflow. The community does."

At this point, some days after this started, more than seven eight nine ten times as many people expressing an opinion want the change reverted as want to keep it. That's a landslide in any voting system. Even if Jeff were right and everyone else were wrong, it would be wrong to persist with it in the face of this kind of community reaction.

The Stack Exchange sites are not a democracy. One person has the ability and in some cases the duty to override the community. This is the reality of the Stack Exchange sites and pretty much every other site on the 'net. These are Jeff's sites. He's better than most people in his position about giving power to the community, but if you thought this was a collaborative venture, if you thought you were a partner in the process, you were mistaken. Your suggestions may be taken on board, or they may be disregarded and your post expressing them deleted, or locked for months on end preventing you even replying with a comment. Maybe it's possible to have a truly community-run site, but I've never seen one. There are going to be occasions when that one, finally-responsible person has to override the community's will.

But in a place that purports to respect the community, put it first, let it run the show, that one person needs to pick his occasions with extreme care. It's not that this change isn't in the same ballpark as something worth rejecting the community over, it's not even in the same country.

"We don't run Stack Overflow. The community does..."

Those are easy words when you agree with the community, or just mildly disagree. The real test is when you strongly disagree. Persisting with this kind of change in the face of this kind of opposition to it demonstrates tremendous disrespect for the community and puts the lie to those words.

Let's not go there. Instead, say "The community is wrong. I'm right. But even so, I am going to accept the community's decision on this."

  • 1
    It's possible that "run Stack Overflow" means moderating, not designing. Jeff isn't interested in the vote because he's not looking at the immediate impact, he's looking years down the road. I still don't agree with him, but I hope I understand his perspective. Jul 9, 2011 at 19:53
  • 14
    @Mark: Jeff Atwood has at some points taken a very active role in moderating math.SE, often to the consternation of many of the elected moderators and the vocal users of the site. His single-handed interventions resulted in the resignation of one of the elected moderators. I was often surprised, among other things, that the head of a largish enterprise would insist on acting on such a small scale: e.g. policing individual comments. When pressed, he has made it clear that he will do what he wants even if most or all of "the community" is against him. Jul 10, 2011 at 4:02
  • 3
    @TJ: +1. I strongly agree with most of what you say. However I disagree that (let me shift away from the personal) the SE platform gives an unusual amount of control to the user. It seemed that way at first, but in the long run I have never been involved with a website that took such an active interest in directing and micromanaging my actions. Protocols on the SE sites change frquently, without warning, and most often without local explanation: you have to go somewhere else (e.g. here) to understand why the site doesn't work the same way today as it did yesterday. Jul 10, 2011 at 4:07
  • @Pete: Let's not bring the Robin Chapman thing into this discussion (or did you mean someone else?). It's an unrelated point and one where (agreeing with your post on the matter) neither party appears to have behaved well. It's true that SE changes regularly and I'd agree those changes are not well-communicated. In general, I'd say it changes for the better, and it makes a change from systems that just never improve, year-on-year. Jul 10, 2011 at 8:14
  • 7
    Don't let numerics fool you. Numerics don't make right. Hundreds of users were opposed to the changes made to the envelope we used to have. See where we are now. Policies here have never been decided solely by votes. If they were, the top entries in the reqs tab on the front page wouldn't still be there, they'd be implemented straight-away. Singular users have been enough to oppose the Team's solid stances and still get what they want implemented. Claiming that all these voting supports should be what defines how the site is created is a non-starter.
    – Grace Note StaffMod
    Jul 13, 2011 at 18:26
  • 5
    @Grace: And I'm not claiming that. I'm saying that the change is pointless, arbitrary, and unnecessary. Forcing a pointless, arbitrary, and unnecessary change down the throats of the community the face of overwhelming feedback to the contrary is demonstrating a complete and total lack of respect for that community and an unwillingness to examine or question oneself. The numbers just prove how unpopular it is. It was wrong before the first person voted; ignoring the feedback is a second, distinct, error. And a worse one, frankly. Jul 13, 2011 at 21:19
  • @TJ: I was not talking about Robin Chapman. (Actually Robin Chapman is apparently still a moderator...) I was talking about the resignation of Akhil Mathew. Jul 15, 2011 at 7:17
  • @Pete and after reading his resignation the loss seems sad. Without getting into the cause, he seems like a valuable member of the community to lose
    – JockM
    Jul 15, 2011 at 16:50
  • @JockM: yes, Akhil Mathew is, in just about every sense I can think of, an exceptionally good person. But he hasn't left the community -- he just stopped being a moderator. So, although I brought it up as an example of negative consequences of outside intervention, I have to say that in absolute terms it wasn't really a tragedy. Jul 16, 2011 at 19:43
  • @Pete I take your point (and actually knew that), but I would argue that it is still a net loss. Just, perhaps, not as grave as it could have been. As a guy who started and ran a community based website for years I know that any community site is only as good as the leaders in it. They are the moderating force, they set the tone; in short: they lead.
    – JockM
    Jul 16, 2011 at 19:53
  • 2
    @JockM: Sure, when a great moderator is driven to resign on principle it seems clear that something is amiss (which is why I brought it up in the first place). I just didn't want to paint too dire a picture. Jul 16, 2011 at 20:02
  • @PeteL.Clark. Jeff Atwood also made some interesting moderation decisions on EL&U. He eventually realised his mistakes and issued a notpology.
    – TRiG
    Nov 2, 2012 at 14:48
  • "ten times as many people expressing an opinion want the change reverted as want to keep it. That's a landslide in any voting system." Your question has 262 in agreement, and 15 in disagreement. Given there are "millions" of users on Stack, that's no representative of "community". Even with 5K views, it's also, quite specifically, a representative of "Meta community" voting, and the majority of the whole "Stack community" have not even read this.
    – James
    Jun 21, 2015 at 17:37
  • @James: The key part of that quote was "expressing an opinion". You can only measure what you can measure. (And it is, of course, now closer to 20 times as many than to ten...) Jun 21, 2015 at 17:43
  • @T.J.Crowder Yes, one can "only measure what you can measure", but that doesn't make it accurate. And one can make a decision without considering the data because it's limited and so not particularly useful. Jeff obviously believes he did what was best for the site, and that will (likely) include the presumed consensus from those who did not come here and vote. That's only speculation of course, and it does touch towards dictatorship. But perhaps if 1 million users had come here and voted in favour of your proposal, Jeff would have at least reconsidered it
    – James
    Jun 21, 2015 at 17:59

I know all has been said. I know I cannot upvote twice. Maybe a picture helps. Below, @ajax333221 was removed, and also that original author is not listed as this a community wiki post:

Are you mad? Who you're talking to?

That bothers me a lot.

  • 1
    I can see that point, we should special-case CW posts here.
    – balpha StaffMod
    Jun 25, 2012 at 12:44
  • 23
    @balpha. No, you certainly should not special-case CW posts. You should get rid of this entire unnecessary annoying "feature".
    – TRiG
    Dec 1, 2012 at 19:00
  • 5
    ...but, @balpha, no time to add that exception in the past 2 years, or did you change your mind? (I just saw @Robert removed here, getting me this...) Or better yet: maybe it's time to kill that darling and remove the removal code altogether now...!?
    – Arjan
    Jun 30, 2014 at 16:35

I am a mathematics professor who has been very active on both Math Overflow and the newer math.SE. The help that I give to others on the site is often on the professional level: indeed, a substantial proportion of the hundreds of questions I have answered on that site have required my expertise as a post-PhD mathematician. Nor has it been one-sided, not at all: there are many other research mathematicians active on the site, including some of the very top ones in the world. When I talk on this site I talk carefully, and when I listen I listen carefully: there is a lot to learn.

Let us call this latest practice what is is: a form of censorship. As an academic behaving in a de facto professional capacity, my reaction to my writing being censored is what you would expect: a mixture of bemusement, bewilderment and outrage. Certain people and organizations have historically had the right to censor, including academic writing. For the last 100 hundred years or so in the Western world, those who have exercised this right have, usually, given very careful thought to make sure that their reasons for censoring are compelling and necessary. The exceptions to this rule have become infamous and long remembered.

Many people here have remarked that the reasons given for this censorship are very far from compelling. In fact it seems to me that they are not reasons at all but rather quickly boil down to I am the boss, so I can make whatever changes will, in my opinion, improve the site. This is true, but as a justification for censorship in a de facto academic environment? Wow. This is going to be negatively remembered by a lot of people for a long time.

Postscript: Some people have said that the SE sites are "wikis" which means that users should expect that others can and will edit their writing. This is technically true, I suppose, but misses the mark in a lot of ways. First, the SE sites are not like wikipedia (in which I have been active) since on the latter site all users enter with essentially full editorial privileges, whereas on SE these privileges are carefully graduated as the user becomes more experienced and "trusted". Second, on math.SE at least, edits to others' questions beyond the level of spelling, formatting and punctuation are done very sparingly: for instance, when I spot what I am convinced is a mathematical error in someone else's answer I do not edit it but rather leave a comment and wait for the answerer to respond. This is respectful and ultimately more efficient -- in most cases the answerer and I (possibly with the help of other commenters) quickly come to an understanding. When people do edit each others' posts even in rather minor ways they often leave comments saying "I changed X; I hope that is okay with you." If some user repeatedly made stylistic changes in others' answers without telling them, the community would quickly intervene to put a stop to it. Third, these edits are edits to the comments, which are strongly linked to a particular person. An analogous change on a wiki would be to make edits in others' signed comments on a discussion page, which everyone has the ability to do and in all my experience on wikipedia no one was stupid enough to attempt. Fourth, these edits are done automatically and without (public, at least) record: if I edit someone else's answer then -- although it rarely comes to this -- they can simply roll it back: the record of the original is preserved. Here the original writing is being altered irrevocably and without even alerting the writer that this is taking place. Thus it is censorship of a rather craven variety, the kind that the censors hope will not be noticed. (But we have noticed it, of course, to say the least.)

  • 5
    -1. Not really related to the original question, IMHO. Jul 9, 2011 at 1:24
  • 1
    Pete, I agree with you. If you simply leave off the @symbol, I don't believe it will be auto-removed. As a PhD in Statistics, I share a very similar desire to maintain professionalism (when it's warranted) and quirky humor when it's not. :O)
    – M. Tibbits
    Jul 9, 2011 at 3:21
  • 8
    This "leaving errors sit while begging the author to fix them" thing really kinda bugs me. Yes, these sites do aim for a delicate balance between encouraging collaboration and encouraging competition, but... Simply refraining to fix things you know are wrong out of some misguided sense of respect for the original author is no more appropriate here than it is on Wikipedia. (I realize this wasn't the point you were intending to make, but my attention was totally derailed upon reading that section)
    – Shog9
    Jul 9, 2011 at 3:22
  • 1
    @MainMa and yet utterly related to the entire conversation as it has evolved
    – JockM
    Jul 9, 2011 at 6:05
  • 8
    @MainMa: the original question is not a question -- it is a response to a recent change in the SE platform. I wanted to make a similar response, and since mine is, broadly speaking, an assent to T.J. Crowder's response, it seems appropriate to post is as an "answer" rather than as a new "question". In this situation, as I understand it a downvote should mean that you disagree with the opinion I am expressing. Is that the case here: do you not believe that the new policy is a form of censorship and/or that it is problematic? I am interested to know what others think. Jul 9, 2011 at 6:45
  • 8
    @MainMa: How it is not related? I said I wanted the new practice removed, for my reasons. He's saying he wants it removed too, for his reasons. Seems pretty darned related to me. :-) Jul 9, 2011 at 8:19
  • 9
    @Shog9: "and invoking the system is the only good reason to begin your comment with an @" -- That's just crazy. The only reason I begin a comment with an @ is to make clear to someone that I'm talking about a person. Remember that we have lots of people who have usernames that are acronyms or trade names or plain lowercase words, like @unwind and @chaos and @Eclipse and @MSN. The "@" disambiguates between a username and a plain word or proper noun. It's good for promoting clarity. Please stop trying to treat this as a pollution issue.
    – Jason S
    Jul 9, 2011 at 17:40
  • 3
    @Shog9: I have to "give examples" why censoring my writing without telling me hampers my ability to express myself? In 2011?? I'm honestly pretty perplexed by this. As I said, John Milton wrote about this in 1644; I'm not going to do a better job than he did. I am definitely operating under the assumption that the SE team are participating members of modern Western society, so that if you stepped back from the particulars of this issue, you actually understand quite well what I am saying. Right? Jul 9, 2011 at 23:43
  • 2
    @Shog9: My guess is what you actually mean is something like this: "Censorship is a harsh and even somewhat loaded word. When you use it, you imply that something really bad is being done. But this is so minor that you should be able to easily adjust to it, if you notice it at all." To that I say: yes, stripping away things like Hello, Dear X and @Y are quite minor as far as censorship goes. The time to complain about censorship is when it starts up a little bit, not when it gets totally out of hand (I imagine complaints are harder at that point). Jul 9, 2011 at 23:51
  • 4
    @Shog9: Manually editing out greetings is not done on math.SE and certainly not in any other community that I belong to. But again the distinction here is between doing something manually after the fact that leaves a record and doing something automatically before the fact that leaves no record. Just so you know: that someone who is designing code for a webpage thinks s/he can decide in advance which of my writing is "superfluous verbiage" inspires a profoundly negative emotional response in me. As a form of censorship, it is significant to me. Please be aware of that. Jul 9, 2011 at 23:56
  • 3
    @Shog9: it's both. It really does hamper my ability to express myself, and on principle I get upset when you write routines to mess with my writing before anyone can see it. And not just me -- this is centuries of social precedent. I do think it's most important to appreciate that. But since you ask... Jul 10, 2011 at 0:02
  • 4
    When I make a comment to an answer, I could for instance be adding something else mathematically to the answer, or I could be engaging the OP directly. It definitely changes the meaning if you strip away the @username entirely. For instance if I write: "@A: Theorem B holds" then I am drawing Theorem B to the attention of user A, and thereby implying that she doesn't know Theorem B already. If I omit @A, then I am directing it to whoever may be reading the question. If the @A gets stripped away a "Yes, I know Theorem B. Why are you suggesting I don't?" response may be coming. Jul 10, 2011 at 0:06
  • 3
    So, yes, this has effects, including possibly offending someone whom you had no intention of offending. Yes, you can work around this by remembering to omit the @ in this situation (or in other ways that I will not get into here). But now I have to keep in my head some fairly complicated site mechanics just to carry out a mathematical conversation: thus my ability to express myself is actually hampered. Jul 10, 2011 at 0:11
  • 3
    @Shog9: I am all for collaborative editing. This is done between people in a way which is best for them both. Writing programs to change my writing is not collaboration: it is, again, censorship. If SE feels that some censorship is necessary to run the site, that is your right. I'm not telling anyone what to do. I'm explaining why this is incredibly problematic to me and, by implication, to a lot of the high-quality users the site should be courting, not alienating. In the long run other sites roughly similar to SE but without these problematic features will spring up, of course. Jul 10, 2011 at 0:18
  • 5
    @Shog9: Thank you for your help, but I am well aware of workarounds to the "feature". As with the OP, I am not asking a question here. I am expressing a serious frustration with the system. I think by now you've understood well what it is that I find problematic. That's all I wanted: to express myself. Jul 10, 2011 at 0:22

I agree with T.J. on this and will add my own additional reason why this disturbs me: because this means that SO (but really Jeff) is exerting editorial control over the style of what I wrote. If there is something offensive (or such) in my content, I accept that is the responsibility and duty of the moderator.

But when you alter the nature of what I write, simply because you don't like it -- and let us face it, this is what it boils down to. Because I don't think you can point to an objective measure of what constitutes noise and what doesn't. It is based on what you prefer.

And look, it is great to have a coherent vision, it is what shapes the greatest products, and I expect you do it in what you write and how you present SO; however when it extends into editorial control over MY style, that crosses a line with me.

I wasn't entirely joking when I suggested your filters might remove (or add) oxford comma's next. While it seems silly it is hard to see a consistant philosophy that says it is OK to remove greetings (a verbal lubricant that dates back to the mists of time), but not something like the Oxford comma.

Addendum: I am not -- in chicken little like way -- trying to say that the sky is falling and SO is editing for content willy-nilly. What I am trying to say is that "clutter" is a deeply deeply subjective thing; and Jeff hasn't even attempted to explain why greetings and @postowner are clutter. He has simply asserted that they are.

And I do not feel it is appropriate to edit other peoples words for style.

Update: And this is it, my final words on this matter -- and possibly my final contribution to SO:

It seems to be there are such bigger forms of noise to address on this site, than to make an arbitrary decision, for an arbitrary definition of noise (because I still haven't seen you give an objective definition of what counts as noise), and enforce it by fiat over the protestations of the community with nothing more than an assertion that you are right.

You insisted that the redundant "@postowner" should be removed because it is noise, and then say that a simple "postowner," is not. This simply makes no sense. It seems like you are acting like that @ convention started with SO and has a single universal understanding of its purpose and meaning. But it didn't start here. It started on other bulletin boards, and forums, and places like twitter; with the purpose of indicating who you are talking to.

I will reiterate what I said in a comment below: I simply do not believe people enter @postowner with a thought in their mind that it will notify the post owner. They do it to indicate who they are talking to. So in that sense their intention is exactly the same as "postowner,".

I think SO has a bigger problem. More than once I have heard colleagues make the comment that they don't participate in SO because they don't like the community. That they feel it is filled with blunt arrogant responses, and that it doesn't feel welcoming. So they reap the benefits of SO without contributing.

Actions like this one, and removing salutations and thanks, take out the kind of vital social lubricant needed to make SO feel less clinical and more collegial. Each time turning SO into a place for people like @sixlettervariables (I am not picking on you sir, but you expressed your opinions rather definitively, and thus make a good example), and less for people like me.

My example is anecdotal, so I cannot say how large it is, but I urge you to take it seriously. Because if you want So to grow, and thrive in the future, then THIS is the kind of signal v. noise you would be worrying about. The one that gets more people to be active and contributing.

All this experience in contributing has done for me is convince me you flaunt precedent (the use of @postowner on other sites, ignoring the social niceties that help make us more individual and relatable, and how links are styled; to name a few), in favor of views you feel are right, but simply cannot prove. Contributing hasn't been a positive experience for me, and it just makes be want to go back to my previous leach status. Which is what I am likely to do.

Look at the votes on your own answer in this matter. Clearly there are more than a few people who disagree with you on this, but you refuse to acknowledge this this with any kind of meaningful response on this, or respond with any kind of citation that shows that removing @postowner, greetings, or thanks has any meaningful affect on the readability and comprehension of conversations here (because ultimately that is what Signal v Noise is about).

I am simply right about this...

  • 2
    There are excellent reasons to leave the @reply intact, but IMO, this isn't it; the SE sites are wikis, and other people will edit your words. Jul 5, 2011 at 2:27
  • 3
    This is a bit over-dramatic. No one is censoring your speech or putting words in your mouth. The system is just removing some clutter. If the system could also automatically correct comma usage accurately, I don't see how that would be a bad thing either.
    – Brad Mace
    Jul 5, 2011 at 15:09
  • 1
    Define clutter. Define clutter in such a way that you could clearly say greetings are clutter, and oxford commas are not; without simply asserting that. Greetings are a natural part of out speech so editing them out is a form of censorship. You may find it to be an acceptable form of one, but it is still one
    – JockM
    Jul 5, 2011 at 21:54
  • 3
    Greetings and salutations are an eye sore. In my other job I've got to fix your problem no matter how nice or rude you are, why can't SO be the same? So save the time writing "Hello" or "Thanks" and include more code or debug output. Then again, I'm a left brained person who just wants folks to get to the point...
    – user7116
    Jul 5, 2011 at 22:22
  • 1
    @BillyONeal There is a difference between editing for content and editing for style. Fixing typos and/or punctuation is not the same thing as changing the tone or writing style.
    – JockM
    Jul 6, 2011 at 0:37
  • 2
    @JockM: If I changed your variable names on SO, that'd be changing your style. Likewise if I only switched "doesn't" to "does not", that'd be changing your style. If I removed a "please help me" or a "thank you for your time good sirs and ma'ams", that'd be editing out extraneous fluff.
    – user7116
    Jul 6, 2011 at 0:42
  • 3
    @JockM: You didn't address the OP in your answer, "Dear T.J. Crowder", nor did you end it with "Sincerely, JockM" (I'm sure T.J. thought that was pretty rude </sarcasm>). These are equally extraneous. You cannot glean anything from "Hello" or "Thank you!!!!11!1one" which helps you answer their question (besides the obvious, "have you actually read the FAQ before posting on SO?"). Edits, manual or automatic, should increase signal and decrease noise.
    – user7116
    Jul 6, 2011 at 1:16
  • 2
    While I'm not sure I've seen it said quite so succinctly, I think most of the "rulings" on these types of issues are consistent with a "get to the point" philosophy.
    – Brad Mace
    Jul 6, 2011 at 2:33
  • 2
    @sixlettervariables I don't want to enforce a style of mandatory greetings or endings in exactly the same way I don't want it enforced by automatically removing them. I am saying that what is signal and what is noise in these cases has no objective measure. They bother you, clearly, but they don't bother me (nor most people I know). So in that regard I agree with you that "Edits, manual or automatic, should increase signal and decrease noise", and would argue that if you cannot prove that something is signal or noise you shouldn't touch it.
    – JockM
    Jul 6, 2011 at 3:37
  • 1
    You know, those Oxford commas are a bit of an eyesore... ;-) Jul 6, 2011 at 4:59
  • 2
    No, but seriously, within a question things like "Hi Fellow Coders!" may be standard social lube, but they don't increase clarity and thus removing them doesn't decrease it; for me, it's a different category. Removing the @ direction does reduce clarity. Irrespective of the notification system, comments may not be directed at the post owner. There's a difference between "Be careful doing this, it can bite you" and "@postowner: Be careful doing this, it can bite you." To remove my explicit direction when commenting is the thing I believe is wrong, unnecessary, and a waste of time. Jul 6, 2011 at 5:01
  • 1
    @sixlettervariables: Style is Jock's thing. Mine is the reduction of clarity, the unnecessary-ness, the waste of time (they're already having to fix bugs in it, and as any software developer will tell you, adding complexity increases maintenance hassles), and the intrusiveness. SO is wiki-like, I expect people (people) to edit my content to an extent. I don't expect a mindless system to strip out my explicit direction, which I've included for clarity, being (again) marginally smarter than the SE engine. :-) Jul 6, 2011 at 11:58
  • 1
    @T.J. There is more at stake than simple clarity. I very much doubt if you did comprehension tests on posts with greetings and signoffs; and ones without that there would be any difference, and no one is attempting to come back with any kind of proof, just opinion. But removing this social lubricant can go a long way to making SO seem dryer, gruffer, less welcoming, more blunt. I encourage anyone to come back with a link to a study or test (or the like) that shows that removing any of these things has any effect on clarity or comprehension.
    – JockM
    Jul 6, 2011 at 14:55
  • 4
    @JockM, unfortunately in this case the only opinion that matters is Jeffs. His is so tightly held that I'm not even sure proof would sway him, if all it proved is that there's no difference - you would need to prove that the lack of greetings was actively harmful, and there's a good chance that the proof would still be rejected. Jul 11, 2011 at 17:33
  • 2
    @Mark And that is a shame. I personally believe that (to quote Carl Sagan) "extraordinary claims require extraordinary proof." I can understand people who say greetings and thanks are not required. But to turn not required into damaging noise without a shred of any kind of any kind of proof makes no sense. But that's ok it's Jeff's playground, and I don't have to stay...
    – JockM
    Jul 11, 2011 at 17:52

Just a reminder of how this came to be (old timers please correct me if I'm wrong).

In the beginning, there were no comments at all, but when comments were implemented, people intuitively started @ing each other. They also had to F5 like crazy to check for answers, since only the post's author was notified of comments.

Then came notifications, the envelope, the SuperCollider. By popular request (and not without some resistance, IIRC) @ comments were added to notifications.

But people kept forgetting to add @, and complained that it was redundant anyway for a first comment. So notifications were tweaked such that if only one person besides the post's author had commented, he would be notified of new comments as well.

The change that is the subject of this discussion was the 'logical' next step: remove redundant @ when only two people (post author and someone else) were commenting, since both of them get notified anyway.

So what?

Mostly, what I want to emphasise (other people have called this out in comments here), is that originally (at least), using @ wasn't about notifications, it was about indicating who the comment was addressed to. In other words, I think that a valid use case is a 'general' comment on a post which isn't directly addressed to the author, but to all readers. The fact that the author will be notified is not directly relevant in this case.

  • I've made a small correction to the timeline regarding the introduction of post-author notifications - they've been around longer than anything else, since comments were originally intended primarily as a way to communicate with the post author. The recent change to notifications is somewhat more subtle, but as you noted does help to explain the logic behind the redundant @ rules. See here for another overview of how messaging has changed over time.
    – Shog9
    Jul 22, 2011 at 7:08
  • @Shog9, thanks, I did attempt to look up some of the oldest feature requests, but either they're not there, my search-fu is bad, or I gave up too quickly... :)
    – Benjol
    Jul 22, 2011 at 10:39

If the system auto-notifies the post owner, great, but removing @postowner from comments is another matter.

This is messing with the contextual clues we get while reading through comments, and the resulting inconsistencies are subtle and manifold.

Here are examples of just two issues that—even without any of the others—should be sufficient to demonstrate that the feature should be reversed:

  • Like many online communities, ours is made up of individuals with an unlimited variety of made-up names. Many of these names are words that are in common use in the English language. This means that even assuming I remember to type "postowner" instead of "@postowner", thereby ensuring the system will post what I've typed, this is all very well when I'm addressing user Joe, but confusion may arise when I'm addressing user Agree or user Cobblers.

  • The first point is exacerbated when a post already has several comments, meaning the foot of the post bearing the poster's name may not be visible at the same time as my comment. Worse, just as I post my comment, another user posts a comment that arrives before mine. The resulting impression is that my comment, lacking as it does its @postowner text (and possibly starting "Agree" or "Cobblers") is a reply to that last comment, when it is in fact a reply to the post owner's last comment.


One could argue that removing parts of my comment without a clear notice is a violation of the CC 4.0 license from SE's part. Unlike previous versions of the license, marking modifications is mandatory in 4.0: https://wiki.creativecommons.org/wiki/License_Versions#Modifications_and_adaptations_must_be_marked_as_such .

  • 3
    I'd reason it is a trivial modification, and no notice is needed. If you wan't to know who of us is right you'll have to go to court....
    – Luuklag
    Sep 24, 2021 at 8:18
  • 1
    @Luuklag True, ultimately; but it's not hard to construct examples in which this change alters significantly the meaning of the comment: for instance comments that contain a pun with OP's username, or use OP's name in the @-notification as part of a sentence. Sep 24, 2021 at 14:46

The @Felix was stripped out automatically because it isn't necessary when you leave a comment for the OP.

The assumption is that unless you address your comment to someone else, it must be for the OP.

  • 26
    I know. I think I made it clear in my bug report that I know why it happened. My point is that it's inappropriate. Jul 2, 2011 at 22:48
  • So you agree with what I said ... I wonder who disagreed?
    – pavium
    Jul 2, 2011 at 22:52
  • 13
    I did, because all your original answer did was repeat what I said in the question about why it happened. (Your edit goes a bit further.) Again, I know why. That is not the point. Jul 2, 2011 at 22:56
  • 3
    I disagree, it's inconsistent. It causes confusion to the user when sometimes the @username is removed and other times it isn't.
    – codewaggle
    Jun 18, 2012 at 11:39

We remove @postowner for the same reason we automatically remove salutations from posts, like

Hi Fellow Coders,

(Note that this is only removed when the only people talking in comments are you and the post owner.)

The @postowner is unnecessary because the post owner is always notified of every comment in every circumstance.

Adding unnecessary name references in comments is yet more words on the page that we have to ignore -- exactly like unnecessary salutations -- so it is ultimately harming the signal-to-noise ratio on the comments.

  • 21
    I disagree, it's inconsistent. It causes confusion to the user when sometimes the @username is removed and other times it isn't. I just saw your comment about not needing it when replying to the postowner, what if you're the postowner replying to a comment?
    – codewaggle
    Jun 18, 2012 at 11:40
  • 21
    The trouble is people get weirded out by having their comments changed for them. Did the ping go through? Did I do something wrong? Should I leave them a note somewhere else, like on a different post or in chat? Is it a bug? Mar 22, 2013 at 20:43
  • 16
    I disagree. Sometimes the comment is addressed to particular person and without the @ reference it doesn't make sense. So actually removing it harms readability of the page.
    – Tomas
    Jun 26, 2013 at 14:17
  • 4
    I tend to agree with Jeff. @othercommenter makes sense, but postowner doesn't.
    – jsnfwlr
    Aug 22, 2013 at 3:34

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .