Slashdot Mirror

A few weeks ago you had a chance to ask author, entrepreneur, and software developer Jeff Atwood about founding Stack Overflow and the Stack Exchange Network, as well as his new endeavor, the Discourse open-source discussion platform. Below you will find his answers to your questions. Magic wand
by Anonymous Coward

If you had a magic wand to make one change in technology right now, what would it be?

Atwood: Users would not have to generate, remember, enter, or ever think about passwords again. Computers would automatically know who the user is through a combination of ambient biometrics plus physical possession of some kind of device. Like, say, a smartphone.

Passwords are the enemy. And the users, because we are the idiots put in charge of making up the passwords. But mostly, it's the *goddamn passwords*.



Why did you choose Microsoft Platform for SE?
by Sadsfae

I don't see many large, high profile sites running an entire Microsoft Windows stack nowadays (IIS/SQL Server, etc) but Stack Exchange is one of them.

What were the reasons behind choosing a full Microsoft stack versus any of the Open Source alternatives which seem much more prevalent, especially in start-ups and smaller businesses for web presence?

Atwood: Mostly, C# is what I knew and what I was skilled in -- and I'm a great fan of its primary architect Anders Hejsberg who also created Turbo Pascal and Delphi. Performance was a goal, too, and since C# is a compiled language it's *extremely* fast. I think you can see for yourself that Stack Overflow is absurdly fast. Having switched to Ruby with the Discourse project, I can also testify that Ruby .. is, uh ... not ... absurdly fast.

The only downside of the .NET environment is, honestly, the SQL Server licensing costs which can be quite extreme at scale. There is movement to make .NET more open source. Plus the long running mono effort.

The main weakness of .NET is that it's not great for open source projects, though that has changed a bit over the last few years. It never really made sense to open source Stack Overflow -- ask yourself, how many Stack Overflow clones have flourished? Why is that? As a closed source project, the performance, great language design, and scaling of C# worked for us.

I made different choices for Discourse which *was* designed to be an open source project, a tool that is widely applicable to many communities, from day zero.



History of StackExchange
by unencode200x

A question on the history of Stack Exchange. What was the original idea that drove you to make StackExchange and how has it evolved or added since?

Atwood: Do you remember a site called Experts-Exchange? No? That means we succeeded at our original goal.

The basic concept was to do a 100% community driven Q&A that had elements of:

- Wikipedia (all the articles are always up to date and not dead tombstones from six years ago).
- Reddit (voting up the good information and voting down the bad information).
- Blogs (ownership, curation, and responsibility for content that has your name on it).
- Videogames (the Xbox 360 Achievements system, points, and ways of encouraging and incentivizing positive community behavior that are fun).

Where everything we build together is creative commons, and belongs to all of us, since you guys and gals are the ones doing all the work in the system!



Reputation mechanisms & scientific quality
by Anonymous Coward

Jeff, have you thought about how to use reputation mechanisms to improve the quality of published scientific results? I'm asking in the context of John P. A. Ioannidis' famous paper.

It seems to me one fix for this (horrible) problem might be an online reputation mechanism where scientists could rate the reproducibility of published results. Thoughts? (thanks for inventing Stack Exchange - you've done the world a big favor).

Atwood: It certainly seems applicable. The Stack engine works best for systems of data, fact, and science -- or at least a "tome of knowledge" -- where you can actually verify an answer (or five answers) as plausibly correct. You can see which topics do best on the Stack engine in the Stack Exchange directory, with a massive, Jupiter sized Stack Overflow right at the top.

There's always more than one way to do it, of course, but when you start getting dozens or hundreds of "answers" you don't have Q&A, you have a discussion with no clear answers, just opinions.



User Reputation, Moderating, and Discourse
by T.E.D.

I think its probably inarguable that the biggest innovation StackOverflow brought to the web was the centrality of reputation and user moderation to its design. Sure, our own /. had done something similar years before, and it was hardly the first either, but no website I know of had before taken it to its logical conclusion in quite the way SO does. This effectively "crowdsourced" a lot of traditional website administrative activities, which turned out to be an incredibly powerful idea. Practically all the functionality of SO is built around the concept.

So when I saw you were tackling online message boards, I expected the same kind of thing. But browsing around a typical Discourse thread, I'm not seeing that at all. Sure, users can "heart" posts, but all that does is bump a small counter next to the heart. There is no way to tell at a glance which posts users found the best and/or worst. Higher rated posts don't sort to the top, or get bigger or anything. As a result, I don't even see that feature used much. Certainly its nothing like SO, where post voting is the central activity. It also seems like moderation on Discourse is designed to be done by administrators, not users. I don't see any facility for users getting moderation privs as they gain reputation. Compared to SO, Discourse seems kind of, well, like a big step backwards in interactivity.

I'm sure I'm missing something here. What is it? Or did you really decide SO's centering of its design around users and their opinion on posts was a mistake, or perhaps just not a good fit for a more generalized discussion board?

Atwood: Sorting a conversation by votes is a pretty effective way to destroy conversation. How can you follow the logical flow of back and forth, chronological dialog when the ground is constantly shifting underneath you as posts get voted up or down? You can't.

Stack is a system of technical Q&A, where opinions are fascinating, and all, but they are completely trumped by facts, data, and science. Stack only tolerates the minimum amount of discussion necessary to get the best questions and the best answers. The goal isn't for people to talk to each other, the goal is for people to *answer the damn question*. Ideally with the aforementioned facts, data, and science, so our peers can objectively decide if the answer is correct and works.

Discourse, on the other hand, is explicitly a system of discussion and opinion. There is no right and wrong. You can't tell me my opinion that Wolverine is the coolest X-Man is wrong. Long after people have forgotten what exactly was said, they will remember how you made them feel. That's what the like (heart) action is about, and why it is featured so prominently: empathy. Discourse is a system of empathy.

We do have user trust levels in Discourse, it's just less obvious, because we're playing a different kind of game. Compare that with Stack, where your reputation number and badge counts appear prominently next to your name every time you post. Trust me, people *do* notice when you like their post. And if they see your post got 20 likes whereas their post only got two, or none, they absolutely notice that too. Discourse is more of a collaborative game, where Stack is an explicitly competitive game, and that's why the score is so prominent. The best way to motivate a programmer is to tell them someone else did it better. Don't try to race sheep, don't try to herd race horses.

There may not be re-ordering by votes in Discourse, as there is in Stack Overflow, but there is a summary mode for topics when they reach 50 or more replies. If you'd like to see this in action, visit a longer topic and press the "Summarize Topic" button at the top near the estimated read time, or as I like to call it, the TL;DR button. Then you'll see only the "best" 10 percent of that long topic. That factors in a lot of data from each post such as likes, incoming and outgoing links to the post, the number of times it has been read, total read time (we track actual on-screen read time for all posts in Discourse), number of times it has been bookmarked, number of replies, and so on. You can also expand context above (in reply to) and below (replies) for each post as needed. Or you can expand the collapsed gaps as needed. Try it, you just need a long topic with over 50 replies to see it in action.



Cargo cult programming and Stack Overflow
by Anonymous Coward

I don't mean to minimize StackOverflow's contribution to the online knowledge base, because it's a great tool when used properly. I'm a systems guy and Server Fault is often more useful than vendor support for looking up strange error messages and possible troubleshooting routes. But, there are a lot of low skill programmers and sysadmins out there who lean on these tools way too much. How do you feel about these properties contributing to the crappy cargo cult programming and sysadmin work we see in our field?

Atwood: Stack is a system of peer education at its core. The key insight is realizing that crappy programmers hurt all of us, and it's our job to learn from each other so that we have less terrible code and terrible coders to deal with in the future. Even if that terrible coder is us!

The best way to learn a topic is to teach it to someone else. That is the skill at the heart of the Stack engine. It works at three levels:

1. Selfish. I need the answer to this question or I may get fired. Give me the answer. This is ideally handled through a good search result. They get what they need.

2. Self Improvement. I want to get reputation and prove to my peers that I know what I am doing. The more I learn, the better I am at my job, the more skilled I am, and the more job satisfaction / money / prestige I can gain.

3. Advancement of our craft. Programming, physics, and math will be here long after we are all dead. It's an honor to help move science forward together as much as we can together in our lifetime, so that future programmers, physicists, and mathematicians can stand on our shoulders and do amazing, incredible things for the future of humanity.

It's fine for people to play the game at level one, because they are also helping others learn and work their way up the skill ladder. If someone is learning, and someone is teaching, we all win.



Rampant closure of questions
by WaffleMonster

From time to time I search stackoverflow for easy answers and I would say about 20% of the time the question has been closed even though it is the reason I went to stackoverflow in the first place. In most of these instances a useful answer was also provided before closure. So my question to you is simply what gives.

The most common reason for closure I run into is that the people closing it don't have any domain clue what is being asked and appear to assume if they don't understand nobody else does either.

Another common reason for closure is the "duplicate" question meme in which nuance is overlooked and questions are marked as duplicates because the people doing the marking failed to understand or appreciate the difference. This is very annoying.

Less common but equally annoying issues are closure due to chatter about domain specific algorithms not being "programming questions" or even more amusing someone posting a question that is more specifically addressed by one of a hundred different stack exchanges even though it is still on topic.

Atwood: Remember that Stack is for questions that can be explicitly answered, *not* discussion. It's not a place for "what's the best way to.." opinion sinkholes. Humans love this kind of stuff because they are social animals, and there's nothing wrong with wanting to have a discussion -- you just need to have that discussion elsewhere, because you can't have it on Stack Overflow.

We're strict about this because we've seen what happens when systems are not explicit about their goals. This way lies madness. This way lies Yahoo Answers. In Stack's case the goal is *learning*. And I do not mean accidental, random, meandering, oh-hey-check-out-this-crazy-thing-I-saw-on-Reddit learning, but highly efficient, directed learning where you are a classroom, not a social club.

Duplicates are a hard problem, for sure. That's the one place I felt we didn't make a ton of progress in my four years there, unfortunately. Human beings have the incredibly annoying habit of asking the same questions using completely and utterly different words. They're really good at it. And it is true that over time there are more and more questions and answers on Stack Overflow, so the minefield of "is this a duplicate of...?" is only getting bigger over time. It is a super hard problem. If you have specific ideas of how to handle duplicates better, don't hesitate to ask or search some of the existing topics there.

Remember, Stack Overflow is governed by programmers just like you, you are a citizen of its community just like me, and you get a say in what happens there. You could even be elected a moderator by participating in the yearly elections. So if you don't like the way things work, it's like any other democracy -- make your voice heard, vote, campaign, or run for office.



Relevance of old answers
by Scottingham

As SO ages, some of the offered solutions are no longer valid. Are there currently plans to automate some way of validating old answers automatically? This problem seems to be a larger problem with forums in general. Do you have any musings regarding aging forums?

Atwood: Anyone, even anonymous users, can suggest an edit to any question or answer on any Stack Exchange site. It is like Wikipedia in that regard. Once two other users with sufficient reputation approve the edit it goes through. Alternately, if you earn enough reputation, you can make direct edits yourself.

If you see information that's out of date, edit it to make it more up to date! Be the change you want to see, and all that.



Signal To Noise: Trolls
by Anonymous Coward

In reading your work for years and seeing your various contributions, it seems like you are fascinated with filtering out the most useful information. In many of your blog posts the insight is not yours but rather a conglomeration of chosen useful quotes and sources. I very much appreciate this. My question for you is how do you handle critical feedback vs trolls when dealing with communities. For example, the down button is often a disagree button rather than a negative point. How do you deal with mixed opinions?

To use a real life personal example, TEF noted how he felt you were suggesting that people shouldn't play around to learn. Yet, the way he said it was clearly inflammatory. How do you separate the legitimate concern and critical feedback from the troll who doesn't want to listen to your response?

Atwood: There is a reason we don't have a "dislike" or "downvote" in Discourse. How can an opinion be wrong? It can be rude, offensive, misinformed, misguided, or just plumb crazy -- but it can't be objectively *wrong*. Often the way you judge posts in Discourse is by their *lack* of likes. If nobody feels strongly enough about your reply to push the heart button on it, and 'co-sign' it with their name in public, that says something.

As for separating legitimate criticism from trolling, I don't know that actual trolls, by the strict definition of the word troll, are that common. I think bad faith makes itself quite clear in the tone and delivery of the criticism. Are you saying this in the hope that we can both learn something from the interaction, or are you saying it because you want to hurt or shame or denigrate or discredit me? Truth alone, as it turns out, isn't the whole truth. How you say something matters.

Bad faith is especially visible to the audience, who has no stake in the argument, and can be surprisingly objective in judging authenticity. One of the most striking things about the early days of Stack Overflow was seeing how people would not upvote cruelty. They wouldn't necessarily downvote it, mind you, but overt cruelty and meanness in an answer was never an effective way to get upvotes and reputation even if the technical information was sound. The best way to win an argument isn't to convince the other person, necessarily -- good luck with that -- but to convince everyone who is watching. And you will never convince an audience watching you be cruel to another person.

In life, being cruel to others may achieve some short term goals, but it is *never* a winning long term strategy. And I think that is exactly how it should be.



How do you have a good debate online?
by AmiMoJo

It seems like the internet is mostly a terrible place to have debates. Many forums quickly become echo chambers for people who want to be as offensive as possible just to prove that they can exercise their free speech rights. Other times debates are derailed by cheap tactics like being deliberately offensive to derail the arguments and bog everyone down in accusations that they are "SJWs". Ad-hominems and obvious logical fallacies seem to be the norm.

How do you plan to avoid this happening? So far no-one seems to have found a way.

Atwood: We have a few tricks up our sleeve at Discourse. We try to teach communities not just how the software works, but how human beings should work, with stuff like our Universal Rules of Civilized Discourse which is prominently featured in every install of Discourse and of course Creative Commons licensed.

I believe in the “Just in Time" theory of human behavior where we try to reach you at the exact moment you start typing your first post with the TL;DR version of those rules. And the most important rule of all for empathy is a simple one: hey, there's another person on the other side of that screen. Not just an abstract name, an avatar, a collection of pixels, but a real live human being, just like you.

I don't know if we can weaponize empathy but I'd sure like to try. Not every space has to be open to everyone, unless you work for Facebook or Twitter. Sometimes the point of having your own community is being able to close the door on people who demonstrate that they can't behave themselves in your house. Live with wolves, and you learn to howl.



Stackoverflow in hindsight
by jez9999

In hindsight, would you have reduced the scope of on-topic questions for Stackoverflow to where it's at today when you started the site knowing what you do now, and do you think it would've made the site less popular?

Atwood: Much of the strictness of Stack Overflow evolved as a side effect of the reputation system. Once you have a reputation score, you want to protect that score, and nothing devalues your own reputation more than seeing some other programmer get 300 reputation points from a humorous answer containing nothing but an XKCD comic. Anyone can post an XKCD comic; that takes no particular competence or knowledge. So the evolution in strictness -- peer reputation should come from expression of *skill*, not funny anecdotes -- was largely driven by the community, not by anyone employed at Stack Overflow.

Even knowing what I already knew, which is that putting a number next to someone's name will cause them to do whatever it takes to make that number go up -- I didn't anticipate how strong this effect would be. But ultimately I agree with it, and I think systems should trend to slightly increasing strictness over time as they grow bigger. Big cities have different problems than small cities, and they need more structure.

A few weeks ago you had a chance to ask author, entrepreneur, and software developer Jeff Atwood about founding Stack Overflow and the Stack Exchange Network, as well as his new endeavor, the Discourse open-source discussion platform. Below you will find his answers to your questions. Magic wand
by Anonymous Coward

If you had a magic wand to make one change in technology right now, what would it be?

Atwood: Users would not have to generate, remember, enter, or ever think about passwords again. Computers would automatically know who the user is through a combination of ambient biometrics plus physical possession of some kind of device. Like, say, a smartphone.

Passwords are the enemy. And the users, because we are the idiots put in charge of making up the passwords. But mostly, it's the *goddamn passwords*.



Why did you choose Microsoft Platform for SE?
by Sadsfae

I don't see many large, high profile sites running an entire Microsoft Windows stack nowadays (IIS/SQL Server, etc) but Stack Exchange is one of them.

What were the reasons behind choosing a full Microsoft stack versus any of the Open Source alternatives which seem much more prevalent, especially in start-ups and smaller businesses for web presence?

Atwood: Mostly, C# is what I knew and what I was skilled in -- and I'm a great fan of its primary architect Anders Hejsberg who also created Turbo Pascal and Delphi. Performance was a goal, too, and since C# is a compiled language it's *extremely* fast. I think you can see for yourself that Stack Overflow is absurdly fast. Having switched to Ruby with the Discourse project, I can also testify that Ruby .. is, uh ... not ... absurdly fast.

The only downside of the .NET environment is, honestly, the SQL Server licensing costs which can be quite extreme at scale. There is movement to make .NET more open source. Plus the long running mono effort.

The main weakness of .NET is that it's not great for open source projects, though that has changed a bit over the last few years. It never really made sense to open source Stack Overflow -- ask yourself, how many Stack Overflow clones have flourished? Why is that? As a closed source project, the performance, great language design, and scaling of C# worked for us.

I made different choices for Discourse which *was* designed to be an open source project, a tool that is widely applicable to many communities, from day zero.



History of StackExchange
by unencode200x

A question on the history of Stack Exchange. What was the original idea that drove you to make StackExchange and how has it evolved or added since?

Atwood: Do you remember a site called Experts-Exchange? No? That means we succeeded at our original goal.

The basic concept was to do a 100% community driven Q&A that had elements of:

- Wikipedia (all the articles are always up to date and not dead tombstones from six years ago).
- Reddit (voting up the good information and voting down the bad information).
- Blogs (ownership, curation, and responsibility for content that has your name on it).
- Videogames (the Xbox 360 Achievements system, points, and ways of encouraging and incentivizing positive community behavior that are fun).

Where everything we build together is creative commons, and belongs to all of us, since you guys and gals are the ones doing all the work in the system!



Reputation mechanisms & scientific quality
by Anonymous Coward

Jeff, have you thought about how to use reputation mechanisms to improve the quality of published scientific results? I'm asking in the context of John P. A. Ioannidis' famous paper.

It seems to me one fix for this (horrible) problem might be an online reputation mechanism where scientists could rate the reproducibility of published results. Thoughts? (thanks for inventing Stack Exchange - you've done the world a big favor).

Atwood: It certainly seems applicable. The Stack engine works best for systems of data, fact, and science -- or at least a "tome of knowledge" -- where you can actually verify an answer (or five answers) as plausibly correct. You can see which topics do best on the Stack engine in the Stack Exchange directory, with a massive, Jupiter sized Stack Overflow right at the top.

There's always more than one way to do it, of course, but when you start getting dozens or hundreds of "answers" you don't have Q&A, you have a discussion with no clear answers, just opinions.



User Reputation, Moderating, and Discourse
by T.E.D.

I think its probably inarguable that the biggest innovation StackOverflow brought to the web was the centrality of reputation and user moderation to its design. Sure, our own /. had done something similar years before, and it was hardly the first either, but no website I know of had before taken it to its logical conclusion in quite the way SO does. This effectively "crowdsourced" a lot of traditional website administrative activities, which turned out to be an incredibly powerful idea. Practically all the functionality of SO is built around the concept.

So when I saw you were tackling online message boards, I expected the same kind of thing. But browsing around a typical Discourse thread, I'm not seeing that at all. Sure, users can "heart" posts, but all that does is bump a small counter next to the heart. There is no way to tell at a glance which posts users found the best and/or worst. Higher rated posts don't sort to the top, or get bigger or anything. As a result, I don't even see that feature used much. Certainly its nothing like SO, where post voting is the central activity. It also seems like moderation on Discourse is designed to be done by administrators, not users. I don't see any facility for users getting moderation privs as they gain reputation. Compared to SO, Discourse seems kind of, well, like a big step backwards in interactivity.

I'm sure I'm missing something here. What is it? Or did you really decide SO's centering of its design around users and their opinion on posts was a mistake, or perhaps just not a good fit for a more generalized discussion board?

Atwood: Sorting a conversation by votes is a pretty effective way to destroy conversation. How can you follow the logical flow of back and forth, chronological dialog when the ground is constantly shifting underneath you as posts get voted up or down? You can't.

Stack is a system of technical Q&A, where opinions are fascinating, and all, but they are completely trumped by facts, data, and science. Stack only tolerates the minimum amount of discussion necessary to get the best questions and the best answers. The goal isn't for people to talk to each other, the goal is for people to *answer the damn question*. Ideally with the aforementioned facts, data, and science, so our peers can objectively decide if the answer is correct and works.

Discourse, on the other hand, is explicitly a system of discussion and opinion. There is no right and wrong. You can't tell me my opinion that Wolverine is the coolest X-Man is wrong. Long after people have forgotten what exactly was said, they will remember how you made them feel. That's what the like (heart) action is about, and why it is featured so prominently: empathy. Discourse is a system of empathy.

We do have user trust levels in Discourse, it's just less obvious, because we're playing a different kind of game. Compare that with Stack, where your reputation number and badge counts appear prominently next to your name every time you post. Trust me, people *do* notice when you like their post. And if they see your post got 20 likes whereas their post only got two, or none, they absolutely notice that too. Discourse is more of a collaborative game, where Stack is an explicitly competitive game, and that's why the score is so prominent. The best way to motivate a programmer is to tell them someone else did it better. Don't try to race sheep, don't try to herd race horses.

There may not be re-ordering by votes in Discourse, as there is in Stack Overflow, but there is a summary mode for topics when they reach 50 or more replies. If you'd like to see this in action, visit a longer topic and press the "Summarize Topic" button at the top near the estimated read time, or as I like to call it, the TL;DR button. Then you'll see only the "best" 10 percent of that long topic. That factors in a lot of data from each post such as likes, incoming and outgoing links to the post, the number of times it has been read, total read time (we track actual on-screen read time for all posts in Discourse), number of times it has been bookmarked, number of replies, and so on. You can also expand context above (in reply to) and below (replies) for each post as needed. Or you can expand the collapsed gaps as needed. Try it, you just need a long topic with over 50 replies to see it in action.



Cargo cult programming and Stack Overflow
by Anonymous Coward

I don't mean to minimize StackOverflow's contribution to the online knowledge base, because it's a great tool when used properly. I'm a systems guy and Server Fault is often more useful than vendor support for looking up strange error messages and possible troubleshooting routes. But, there are a lot of low skill programmers and sysadmins out there who lean on these tools way too much. How do you feel about these properties contributing to the crappy cargo cult programming and sysadmin work we see in our field?

Atwood: Stack is a system of peer education at its core. The key insight is realizing that crappy programmers hurt all of us, and it's our job to learn from each other so that we have less terrible code and terrible coders to deal with in the future. Even if that terrible coder is us!

The best way to learn a topic is to teach it to someone else. That is the skill at the heart of the Stack engine. It works at three levels:

1. Selfish. I need the answer to this question or I may get fired. Give me the answer. This is ideally handled through a good search result. They get what they need.

2. Self Improvement. I want to get reputation and prove to my peers that I know what I am doing. The more I learn, the better I am at my job, the more skilled I am, and the more job satisfaction / money / prestige I can gain.

3. Advancement of our craft. Programming, physics, and math will be here long after we are all dead. It's an honor to help move science forward together as much as we can together in our lifetime, so that future programmers, physicists, and mathematicians can stand on our shoulders and do amazing, incredible things for the future of humanity.

It's fine for people to play the game at level one, because they are also helping others learn and work their way up the skill ladder. If someone is learning, and someone is teaching, we all win.



Rampant closure of questions
by WaffleMonster

From time to time I search stackoverflow for easy answers and I would say about 20% of the time the question has been closed even though it is the reason I went to stackoverflow in the first place. In most of these instances a useful answer was also provided before closure. So my question to you is simply what gives.

The most common reason for closure I run into is that the people closing it don't have any domain clue what is being asked and appear to assume if they don't understand nobody else does either.

Another common reason for closure is the "duplicate" question meme in which nuance is overlooked and questions are marked as duplicates because the people doing the marking failed to understand or appreciate the difference. This is very annoying.

Less common but equally annoying issues are closure due to chatter about domain specific algorithms not being "programming questions" or even more amusing someone posting a question that is more specifically addressed by one of a hundred different stack exchanges even though it is still on topic.

Atwood: Remember that Stack is for questions that can be explicitly answered, *not* discussion. It's not a place for "what's the best way to.." opinion sinkholes. Humans love this kind of stuff because they are social animals, and there's nothing wrong with wanting to have a discussion -- you just need to have that discussion elsewhere, because you can't have it on Stack Overflow.

We're strict about this because we've seen what happens when systems are not explicit about their goals. This way lies madness. This way lies Yahoo Answers. In Stack's case the goal is *learning*. And I do not mean accidental, random, meandering, oh-hey-check-out-this-crazy-thing-I-saw-on-Reddit learning, but highly efficient, directed learning where you are a classroom, not a social club.

Duplicates are a hard problem, for sure. That's the one place I felt we didn't make a ton of progress in my four years there, unfortunately. Human beings have the incredibly annoying habit of asking the same questions using completely and utterly different words. They're really good at it. And it is true that over time there are more and more questions and answers on Stack Overflow, so the minefield of "is this a duplicate of...?" is only getting bigger over time. It is a super hard problem. If you have specific ideas of how to handle duplicates better, don't hesitate to ask or search some of the existing topics there.

Remember, Stack Overflow is governed by programmers just like you, you are a citizen of its community just like me, and you get a say in what happens there. You could even be elected a moderator by participating in the yearly elections. So if you don't like the way things work, it's like any other democracy -- make your voice heard, vote, campaign, or run for office.



Relevance of old answers
by Scottingham

As SO ages, some of the offered solutions are no longer valid. Are there currently plans to automate some way of validating old answers automatically? This problem seems to be a larger problem with forums in general. Do you have any musings regarding aging forums?

Atwood: Anyone, even anonymous users, can suggest an edit to any question or answer on any Stack Exchange site. It is like Wikipedia in that regard. Once two other users with sufficient reputation approve the edit it goes through. Alternately, if you earn enough reputation, you can make direct edits yourself.

If you see information that's out of date, edit it to make it more up to date! Be the change you want to see, and all that.



Signal To Noise: Trolls
by Anonymous Coward

In reading your work for years and seeing your various contributions, it seems like you are fascinated with filtering out the most useful information. In many of your blog posts the insight is not yours but rather a conglomeration of chosen useful quotes and sources. I very much appreciate this. My question for you is how do you handle critical feedback vs trolls when dealing with communities. For example, the down button is often a disagree button rather than a negative point. How do you deal with mixed opinions?

To use a real life personal example, TEF noted how he felt you were suggesting that people shouldn't play around to learn. Yet, the way he said it was clearly inflammatory. How do you separate the legitimate concern and critical feedback from the troll who doesn't want to listen to your response?

Atwood: There is a reason we don't have a "dislike" or "downvote" in Discourse. How can an opinion be wrong? It can be rude, offensive, misinformed, misguided, or just plumb crazy -- but it can't be objectively *wrong*. Often the way you judge posts in Discourse is by their *lack* of likes. If nobody feels strongly enough about your reply to push the heart button on it, and 'co-sign' it with their name in public, that says something.

As for separating legitimate criticism from trolling, I don't know that actual trolls, by the strict definition of the word troll, are that common. I think bad faith makes itself quite clear in the tone and delivery of the criticism. Are you saying this in the hope that we can both learn something from the interaction, or are you saying it because you want to hurt or shame or denigrate or discredit me? Truth alone, as it turns out, isn't the whole truth. How you say something matters.

Bad faith is especially visible to the audience, who has no stake in the argument, and can be surprisingly objective in judging authenticity. One of the most striking things about the early days of Stack Overflow was seeing how people would not upvote cruelty. They wouldn't necessarily downvote it, mind you, but overt cruelty and meanness in an answer was never an effective way to get upvotes and reputation even if the technical information was sound. The best way to win an argument isn't to convince the other person, necessarily -- good luck with that -- but to convince everyone who is watching. And you will never convince an audience watching you be cruel to another person.

In life, being cruel to others may achieve some short term goals, but it is *never* a winning long term strategy. And I think that is exactly how it should be.



How do you have a good debate online?
by AmiMoJo

It seems like the internet is mostly a terrible place to have debates. Many forums quickly become echo chambers for people who want to be as offensive as possible just to prove that they can exercise their free speech rights. Other times debates are derailed by cheap tactics like being deliberately offensive to derail the arguments and bog everyone down in accusations that they are "SJWs". Ad-hominems and obvious logical fallacies seem to be the norm.

How do you plan to avoid this happening? So far no-one seems to have found a way.

Atwood: We have a few tricks up our sleeve at Discourse. We try to teach communities not just how the software works, but how human beings should work, with stuff like our Universal Rules of Civilized Discourse which is prominently featured in every install of Discourse and of course Creative Commons licensed.

I believe in the “Just in Time" theory of human behavior where we try to reach you at the exact moment you start typing your first post with the TL;DR version of those rules. And the most important rule of all for empathy is a simple one: hey, there's another person on the other side of that screen. Not just an abstract name, an avatar, a collection of pixels, but a real live human being, just like you.

I don't know if we can weaponize empathy but I'd sure like to try. Not every space has to be open to everyone, unless you work for Facebook or Twitter. Sometimes the point of having your own community is being able to close the door on people who demonstrate that they can't behave themselves in your house. Live with wolves, and you learn to howl.



Stackoverflow in hindsight
by jez9999

In hindsight, would you have reduced the scope of on-topic questions for Stackoverflow to where it's at today when you started the site knowing what you do now, and do you think it would've made the site less popular?

Atwood: Much of the strictness of Stack Overflow evolved as a side effect of the reputation system. Once you have a reputation score, you want to protect that score, and nothing devalues your own reputation more than seeing some other programmer get 300 reputation points from a humorous answer containing nothing but an XKCD comic. Anyone can post an XKCD comic; that takes no particular competence or knowledge. So the evolution in strictness -- peer reputation should come from expression of *skill*, not funny anecdotes -- was largely driven by the community, not by anyone employed at Stack Overflow.

Even knowing what I already knew, which is that putting a number next to someone's name will cause them to do whatever it takes to make that number go up -- I didn't anticipate how strong this effect would be. But ultimately I agree with it, and I think systems should trend to slightly increasing strictness over time as they grow bigger. Big cities have different problems than small cities, and they need more structure.

A few weeks ago you had a chance to ask author, entrepreneur, and software developer Jeff Atwood about founding Stack Overflow and the Stack Exchange Network, as well as his new endeavor, the Discourse open-source discussion platform. Below you will find his answers to your questions. Magic wand
by Anonymous Coward

If you had a magic wand to make one change in technology right now, what would it be?

Atwood: Users would not have to generate, remember, enter, or ever think about passwords again. Computers would automatically know who the user is through a combination of ambient biometrics plus physical possession of some kind of device. Like, say, a smartphone.

Passwords are the enemy. And the users, because we are the idiots put in charge of making up the passwords. But mostly, it's the *goddamn passwords*.



Why did you choose Microsoft Platform for SE?
by Sadsfae

I don't see many large, high profile sites running an entire Microsoft Windows stack nowadays (IIS/SQL Server, etc) but Stack Exchange is one of them.

What were the reasons behind choosing a full Microsoft stack versus any of the Open Source alternatives which seem much more prevalent, especially in start-ups and smaller businesses for web presence?

Atwood: Mostly, C# is what I knew and what I was skilled in -- and I'm a great fan of its primary architect Anders Hejsberg who also created Turbo Pascal and Delphi. Performance was a goal, too, and since C# is a compiled language it's *extremely* fast. I think you can see for yourself that Stack Overflow is absurdly fast. Having switched to Ruby with the Discourse project, I can also testify that Ruby .. is, uh ... not ... absurdly fast.

The only downside of the .NET environment is, honestly, the SQL Server licensing costs which can be quite extreme at scale. There is movement to make .NET more open source. Plus the long running mono effort.

The main weakness of .NET is that it's not great for open source projects, though that has changed a bit over the last few years. It never really made sense to open source Stack Overflow -- ask yourself, how many Stack Overflow clones have flourished? Why is that? As a closed source project, the performance, great language design, and scaling of C# worked for us.

I made different choices for Discourse which *was* designed to be an open source project, a tool that is widely applicable to many communities, from day zero.



History of StackExchange
by unencode200x

A question on the history of Stack Exchange. What was the original idea that drove you to make StackExchange and how has it evolved or added since?

Atwood: Do you remember a site called Experts-Exchange? No? That means we succeeded at our original goal.

The basic concept was to do a 100% community driven Q&A that had elements of:

- Wikipedia (all the articles are always up to date and not dead tombstones from six years ago).
- Reddit (voting up the good information and voting down the bad information).
- Blogs (ownership, curation, and responsibility for content that has your name on it).
- Videogames (the Xbox 360 Achievements system, points, and ways of encouraging and incentivizing positive community behavior that are fun).

Where everything we build together is creative commons, and belongs to all of us, since you guys and gals are the ones doing all the work in the system!



Reputation mechanisms & scientific quality
by Anonymous Coward

Jeff, have you thought about how to use reputation mechanisms to improve the quality of published scientific results? I'm asking in the context of John P. A. Ioannidis' famous paper.

It seems to me one fix for this (horrible) problem might be an online reputation mechanism where scientists could rate the reproducibility of published results. Thoughts? (thanks for inventing Stack Exchange - you've done the world a big favor).

Atwood: It certainly seems applicable. The Stack engine works best for systems of data, fact, and science -- or at least a "tome of knowledge" -- where you can actually verify an answer (or five answers) as plausibly correct. You can see which topics do best on the Stack engine in the Stack Exchange directory, with a massive, Jupiter sized Stack Overflow right at the top.

There's always more than one way to do it, of course, but when you start getting dozens or hundreds of "answers" you don't have Q&A, you have a discussion with no clear answers, just opinions.



User Reputation, Moderating, and Discourse
by T.E.D.

I think its probably inarguable that the biggest innovation StackOverflow brought to the web was the centrality of reputation and user moderation to its design. Sure, our own /. had done something similar years before, and it was hardly the first either, but no website I know of had before taken it to its logical conclusion in quite the way SO does. This effectively "crowdsourced" a lot of traditional website administrative activities, which turned out to be an incredibly powerful idea. Practically all the functionality of SO is built around the concept.

So when I saw you were tackling online message boards, I expected the same kind of thing. But browsing around a typical Discourse thread, I'm not seeing that at all. Sure, users can "heart" posts, but all that does is bump a small counter next to the heart. There is no way to tell at a glance which posts users found the best and/or worst. Higher rated posts don't sort to the top, or get bigger or anything. As a result, I don't even see that feature used much. Certainly its nothing like SO, where post voting is the central activity. It also seems like moderation on Discourse is designed to be done by administrators, not users. I don't see any facility for users getting moderation privs as they gain reputation. Compared to SO, Discourse seems kind of, well, like a big step backwards in interactivity.

I'm sure I'm missing something here. What is it? Or did you really decide SO's centering of its design around users and their opinion on posts was a mistake, or perhaps just not a good fit for a more generalized discussion board?

Atwood: Sorting a conversation by votes is a pretty effective way to destroy conversation. How can you follow the logical flow of back and forth, chronological dialog when the ground is constantly shifting underneath you as posts get voted up or down? You can't.

Stack is a system of technical Q&A, where opinions are fascinating, and all, but they are completely trumped by facts, data, and science. Stack only tolerates the minimum amount of discussion necessary to get the best questions and the best answers. The goal isn't for people to talk to each other, the goal is for people to *answer the damn question*. Ideally with the aforementioned facts, data, and science, so our peers can objectively decide if the answer is correct and works.

Discourse, on the other hand, is explicitly a system of discussion and opinion. There is no right and wrong. You can't tell me my opinion that Wolverine is the coolest X-Man is wrong. Long after people have forgotten what exactly was said, they will remember how you made them feel. That's what the like (heart) action is about, and why it is featured so prominently: empathy. Discourse is a system of empathy.

We do have user trust levels in Discourse, it's just less obvious, because we're playing a different kind of game. Compare that with Stack, where your reputation number and badge counts appear prominently next to your name every time you post. Trust me, people *do* notice when you like their post. And if they see your post got 20 likes whereas their post only got two, or none, they absolutely notice that too. Discourse is more of a collaborative game, where Stack is an explicitly competitive game, and that's why the score is so prominent. The best way to motivate a programmer is to tell them someone else did it better. Don't try to race sheep, don't try to herd race horses.

There may not be re-ordering by votes in Discourse, as there is in Stack Overflow, but there is a summary mode for topics when they reach 50 or more replies. If you'd like to see this in action, visit a longer topic and press the "Summarize Topic" button at the top near the estimated read time, or as I like to call it, the TL;DR button. Then you'll see only the "best" 10 percent of that long topic. That factors in a lot of data from each post such as likes, incoming and outgoing links to the post, the number of times it has been read, total read time (we track actual on-screen read time for all posts in Discourse), number of times it has been bookmarked, number of replies, and so on. You can also expand context above (in reply to) and below (replies) for each post as needed. Or you can expand the collapsed gaps as needed. Try it, you just need a long topic with over 50 replies to see it in action.



Cargo cult programming and Stack Overflow
by Anonymous Coward

I don't mean to minimize StackOverflow's contribution to the online knowledge base, because it's a great tool when used properly. I'm a systems guy and Server Fault is often more useful than vendor support for looking up strange error messages and possible troubleshooting routes. But, there are a lot of low skill programmers and sysadmins out there who lean on these tools way too much. How do you feel about these properties contributing to the crappy cargo cult programming and sysadmin work we see in our field?

Atwood: Stack is a system of peer education at its core. The key insight is realizing that crappy programmers hurt all of us, and it's our job to learn from each other so that we have less terrible code and terrible coders to deal with in the future. Even if that terrible coder is us!

The best way to learn a topic is to teach it to someone else. That is the skill at the heart of the Stack engine. It works at three levels:

1. Selfish. I need the answer to this question or I may get fired. Give me the answer. This is ideally handled through a good search result. They get what they need.

2. Self Improvement. I want to get reputation and prove to my peers that I know what I am doing. The more I learn, the better I am at my job, the more skilled I am, and the more job satisfaction / money / prestige I can gain.

3. Advancement of our craft. Programming, physics, and math will be here long after we are all dead. It's an honor to help move science forward together as much as we can together in our lifetime, so that future programmers, physicists, and mathematicians can stand on our shoulders and do amazing, incredible things for the future of humanity.

It's fine for people to play the game at level one, because they are also helping others learn and work their way up the skill ladder. If someone is learning, and someone is teaching, we all win.



Rampant closure of questions
by WaffleMonster

From time to time I search stackoverflow for easy answers and I would say about 20% of the time the question has been closed even though it is the reason I went to stackoverflow in the first place. In most of these instances a useful answer was also provided before closure. So my question to you is simply what gives.

The most common reason for closure I run into is that the people closing it don't have any domain clue what is being asked and appear to assume if they don't understand nobody else does either.

Another common reason for closure is the "duplicate" question meme in which nuance is overlooked and questions are marked as duplicates because the people doing the marking failed to understand or appreciate the difference. This is very annoying.

Less common but equally annoying issues are closure due to chatter about domain specific algorithms not being "programming questions" or even more amusing someone posting a question that is more specifically addressed by one of a hundred different stack exchanges even though it is still on topic.

Atwood: Remember that Stack is for questions that can be explicitly answered, *not* discussion. It's not a place for "what's the best way to.." opinion sinkholes. Humans love this kind of stuff because they are social animals, and there's nothing wrong with wanting to have a discussion -- you just need to have that discussion elsewhere, because you can't have it on Stack Overflow.

We're strict about this because we've seen what happens when systems are not explicit about their goals. This way lies madness. This way lies Yahoo Answers. In Stack's case the goal is *learning*. And I do not mean accidental, random, meandering, oh-hey-check-out-this-crazy-thing-I-saw-on-Reddit learning, but highly efficient, directed learning where you are a classroom, not a social club.

Duplicates are a hard problem, for sure. That's the one place I felt we didn't make a ton of progress in my four years there, unfortunately. Human beings have the incredibly annoying habit of asking the same questions using completely and utterly different words. They're really good at it. And it is true that over time there are more and more questions and answers on Stack Overflow, so the minefield of "is this a duplicate of...?" is only getting bigger over time. It is a super hard problem. If you have specific ideas of how to handle duplicates better, don't hesitate to ask or search some of the existing topics there.

Remember, Stack Overflow is governed by programmers just like you, you are a citizen of its community just like me, and you get a say in what happens there. You could even be elected a moderator by participating in the yearly elections. So if you don't like the way things work, it's like any other democracy -- make your voice heard, vote, campaign, or run for office.



Relevance of old answers
by Scottingham

As SO ages, some of the offered solutions are no longer valid. Are there currently plans to automate some way of validating old answers automatically? This problem seems to be a larger problem with forums in general. Do you have any musings regarding aging forums?

Atwood: Anyone, even anonymous users, can suggest an edit to any question or answer on any Stack Exchange site. It is like Wikipedia in that regard. Once two other users with sufficient reputation approve the edit it goes through. Alternately, if you earn enough reputation, you can make direct edits yourself.

If you see information that's out of date, edit it to make it more up to date! Be the change you want to see, and all that.



Signal To Noise: Trolls
by Anonymous Coward

In reading your work for years and seeing your various contributions, it seems like you are fascinated with filtering out the most useful information. In many of your blog posts the insight is not yours but rather a conglomeration of chosen useful quotes and sources. I very much appreciate this. My question for you is how do you handle critical feedback vs trolls when dealing with communities. For example, the down button is often a disagree button rather than a negative point. How do you deal with mixed opinions?

To use a real life personal example, TEF noted how he felt you were suggesting that people shouldn't play around to learn. Yet, the way he said it was clearly inflammatory. How do you separate the legitimate concern and critical feedback from the troll who doesn't want to listen to your response?

Atwood: There is a reason we don't have a "dislike" or "downvote" in Discourse. How can an opinion be wrong? It can be rude, offensive, misinformed, misguided, or just plumb crazy -- but it can't be objectively *wrong*. Often the way you judge posts in Discourse is by their *lack* of likes. If nobody feels strongly enough about your reply to push the heart button on it, and 'co-sign' it with their name in public, that says something.

As for separating legitimate criticism from trolling, I don't know that actual trolls, by the strict definition of the word troll, are that common. I think bad faith makes itself quite clear in the tone and delivery of the criticism. Are you saying this in the hope that we can both learn something from the interaction, or are you saying it because you want to hurt or shame or denigrate or discredit me? Truth alone, as it turns out, isn't the whole truth. How you say something matters.

Bad faith is especially visible to the audience, who has no stake in the argument, and can be surprisingly objective in judging authenticity. One of the most striking things about the early days of Stack Overflow was seeing how people would not upvote cruelty. They wouldn't necessarily downvote it, mind you, but overt cruelty and meanness in an answer was never an effective way to get upvotes and reputation even if the technical information was sound. The best way to win an argument isn't to convince the other person, necessarily -- good luck with that -- but to convince everyone who is watching. And you will never convince an audience watching you be cruel to another person.

In life, being cruel to others may achieve some short term goals, but it is *never* a winning long term strategy. And I think that is exactly how it should be.



How do you have a good debate online?
by AmiMoJo

It seems like the internet is mostly a terrible place to have debates. Many forums quickly become echo chambers for people who want to be as offensive as possible just to prove that they can exercise their free speech rights. Other times debates are derailed by cheap tactics like being deliberately offensive to derail the arguments and bog everyone down in accusations that they are "SJWs". Ad-hominems and obvious logical fallacies seem to be the norm.

How do you plan to avoid this happening? So far no-one seems to have found a way.

Atwood: We have a few tricks up our sleeve at Discourse. We try to teach communities not just how the software works, but how human beings should work, with stuff like our Universal Rules of Civilized Discourse which is prominently featured in every install of Discourse and of course Creative Commons licensed.

I believe in the “Just in Time" theory of human behavior where we try to reach you at the exact moment you start typing your first post with the TL;DR version of those rules. And the most important rule of all for empathy is a simple one: hey, there's another person on the other side of that screen. Not just an abstract name, an avatar, a collection of pixels, but a real live human being, just like you.

I don't know if we can weaponize empathy but I'd sure like to try. Not every space has to be open to everyone, unless you work for Facebook or Twitter. Sometimes the point of having your own community is being able to close the door on people who demonstrate that they can't behave themselves in your house. Live with wolves, and you learn to howl.



Stackoverflow in hindsight
by jez9999

In hindsight, would you have reduced the scope of on-topic questions for Stackoverflow to where it's at today when you started the site knowing what you do now, and do you think it would've made the site less popular?

Atwood: Much of the strictness of Stack Overflow evolved as a side effect of the reputation system. Once you have a reputation score, you want to protect that score, and nothing devalues your own reputation more than seeing some other programmer get 300 reputation points from a humorous answer containing nothing but an XKCD comic. Anyone can post an XKCD comic; that takes no particular competence or knowledge. So the evolution in strictness -- peer reputation should come from expression of *skill*, not funny anecdotes -- was largely driven by the community, not by anyone employed at Stack Overflow.

Even knowing what I already knew, which is that putting a number next to someone's name will cause them to do whatever it takes to make that number go up -- I didn't anticipate how strong this effect would be. But ultimately I agree with it, and I think systems should trend to slightly increasing strictness over time as they grow bigger. Big cities have different problems than small cities, and they need more structure.

Jeff Atwood is an author, entrepreneur, and software developer. He runs the popular programming blog Coding Horror and is the co-founder of Stack Overflow and the Stack Exchange Network. In early 2012 he decided to leave Stack Exchange so he could spend more time with his family. A year later he announced his new company the Civilized Discourse Construction Kit, Inc. and the Discourse open-source discussion platform which aims to improve conversations on the internet. Jeff has agreed to give some of his time to answer any questions you may have. As usual, ask as many as you'd like, but please, one question per post.

An anonymous reader writes "Jeff Atwood has a post on his Coding Horror weblog about his latest project, Discourse, 'a next-generation, 100% open source discussion platform built for the next decade of the Internet.' Along with Coding Horror, Jeff is most well-known for his work on Stack Exchange and its family of related sites. In the same way that he tried to improve Q&A sites, he hopes to make forum/discussion software better with a team of folks he's pulled together for the task. They're using the 'Wordpress model' of offering both open source software and commercial offerings. The software interface is an in-browser app via Ember.js, with a Ruby on Rails and Postgres backend. I wonder if it will ever have an NNTP gateway."

An anonymous reader writes "Jeff Atwood has a post on his Coding Horror weblog about his latest project, Discourse, 'a next-generation, 100% open source discussion platform built for the next decade of the Internet.' Along with Coding Horror, Jeff is most well-known for his work on Stack Exchange and its family of related sites. In the same way that he tried to improve Q&A sites, he hopes to make forum/discussion software better with a team of folks he's pulled together for the task. They're using the 'Wordpress model' of offering both open source software and commercial offerings. The software interface is an in-browser app via Ember.js, with a Ruby on Rails and Postgres backend. I wonder if it will ever have an NNTP gateway."