The MIT License – Clarity on Using Code on Stack Overflow and Stack Exchange

Question

Update (Dec. 22, 2015): Thanks, everyone, for your feedback to this proposal. We're going to digest this one over the holidays and should have a follow-up announcement answering your questions and addressing your concerns after the new year. We won't be making any hurried decisions on this topic, and certainly not without ample opportunity to integrate your ideas into the final decision. So continue letting us know what you think. You'll hear again from us soon.

Update (Jan. 14, 2016): The promised follow-up is here, requiring attribution and postponing the change to March 1, 2016.

CC-BY-SA is an ideal license for a crowdsourced knowledge base. We’ve benefited immensely from it, our community has gained protection through it, and we look forward to CC-BY-SA continuing to cover prose contributions to Stack Overflow and across the network for all of eternity.

But code is a bit different (pun intended), and it’s always been a little ambiguous how CC-BY-SA covers code. This has led to uncertainty among conscientious developers as they've struggled to understand what (if anything) the license requires of them when grabbing a few lines of code from a post on Stack Exchange. Uncertainty is a drag on productivity, for you and for us, and we feel obligated to make code use more clear.

Starting Feb 1, 2016, all new code contributions to Stack Overflow and Stack Exchange will be covered by the MIT License.

We’ve been working on this problem for a little while now. We’ve consulted with the OSI, enlisted the help of a real-life lawyer, and discussed it exhaustively internally. We’ve come up with a solution that we think is a major improvement upon the status quo.

The new license terms, in brief:

Starting February 1, 2016, contributions across the network will be licensed to the public under the following terms:

• Non-code contributions will continue to be available for use under the terms of CC-BY-SA
• Code contributions will be available for use under the terms of the MIT License
• You don’t have to include the full MIT License in your code base. Contributors agree to give code users permission to ignore the MIT License’s notice preservation requirement, as long as users give reasonable attribution upon request of the copyright holder (or Stack Exchange on behalf of the contributor). This optional exception to the MIT License will live in our terms of service.

That last bullet makes using Stack Exchange easy, and provides added protection to code contributors and users.

For the vast majority of developers, the above is Too Much Information. The new licensing terms don’t change anything. You can do everything you did before and maybe more when you visit Stack Overflow: copy code, tinker, fiddle, put it in your project, and keep building.

But to future-proof your work, we recommend you do one of these 2 things, or both:

• A) Add a comment to your code that links back to the post where you found it, or
• B) Comply with the MIT as it’s typically used, by including the full license text in your source

You’re doing option A already, right? This is just standard operating procedure when it comes to finding code on the internet – a hyperlink comment ensures you’ll be able to debug down the line. But under the new terms a hyperlink comment is more than just pragmatic, it’s a hat-tip, and it’s a tit-for-tat that ensures you’ve complied with a contributor’s terms.

We want to know what you think about the terms described above. Before you let us know, a few anticipated questions answered in brief:

Is it really that simple?

It’s pretty simple. We’re optimizing for ease of compliance and clarity. This scheme lets you be compliant even if you don’t do a thing. You just have to do a thing – provide credit – when and if the original poster asks for it.

My project is Open Source, can't I just include the MIT License?

Yes. Taking advantage of the notice preservation exception is completely optional. When the new terms launch, we'll have easy to follow guidelines in the help center on how to be compliant for a variety of use cases.

What about code contributed before February 1, 2016?

This is just a first step in alleviating licensing ambiguity for users of Stack Overflow and other Stack Exchange sites. Code that was contributed prior to February 1, 2016 will be available for use as it has always been, under the terms of CC-BY-SA. We are looking into ways we might allow users to opt-in old code contributions to the new scheme.

What about other sites in the Stack Exchange network?

These terms will go into effect for all code posted after Feb. 1 on all sites in the Stack Exchange network.

Who came up with this scheme?

You can blame or optionally thank Tim Post, Britton Payne and Sam Brand for this. They also had an enormous amount of help from the Open Source Initiative.

What’s next?

We hope to roll this out February 1, 2016. We know it’s short notice, but we think the proposed terms are a decisive improvement over status quo, and so a change probably can’t come early enough. Please let us know what you think.

Answer 1

In essence, this proposal is to move from a copyleft license (CC-BY-SA 3.0) to a permissive license for code. (By default, it's basically the most permissive license possible, nearly equivalent to public domain with a liability disclaimer.)

Having thought about it for a couple of hours, I've shifted my opinion from "I'm not sure" to "Please reconsider — this is truly awful." (Sorry to sound dramatic, but @Laura suggests that that's what it will take.)

I suppose that the rationale for the change is to protect users of Stack Overflow answers from (hopefully hypothetical) copyright trolls whose modus operandi is:

1. Post code in a Stack Overflow answer
2. Wait for the code to be incorporated into some software project
3. Sue or extort
4. Win or settle
5. Profit!

To preempt that low-probability, high-stakes terroristic threat, we're being faced with a proposal that imposes huge costs on all Stack Exchange communities.

Wrong attitude

When I contribute an answer on Stack Exchange, I like to think that I am sharing knowledge, not giving away code, even if my answer happens to contain code. Also, we as a community expect questions to be more like "How should I …?" rather than "Can anyone gimme teh codez to do …?" The Creative Commons license expresses the knowledge-sharing attitude. The MIT License most certainly does not.

MIT License: inappropriate for questions

When someone posts a question, whether to Stack Overflow, Code Review, WordPress, Raspberry Pi, or any other Stack Exchange site, I don't believe that there is any intention to donate that code to the public domain (technically not PD, but in practice indistinguishable).

For Code Review, in particular, the code being shared is usually substantial, often a complete runnable program. The intention is to share code for discussion, not to give code away. A permissive license that allows commercial exploitation without attribution could discourage a lot of questions. It could also encourage users to withhold parts of their code, hindering open discussion.

Sometimes questions are actually homework solutions. It would be helpful to have the attribution requirement as an additional legal tool to combat plagiarism, in addition to the usual moral/academic standards.

MIT License: inappropriate for answers

We want the Internet to take the knowledge and advice that they learn from answers to improve their code and their skills. We do not expect them to just take the code as a product.

If the answer is trivial (e.g. "Just use this library function that does exactly what you need"), then common sense says that there is no copyright licensing issue. You just learned something about the obvious way to accomplish something.

If the answer contains a substantial original contribution of ready-to-use code, then attribution should be required by default.

Let's take a hypothetical extreme example: someone publishes an O'Reilly-style cookbook consisting of code snippets extracted from Stack Overflow, Code Review, PPCG, …. Only the code blocks are taken; all of the surrounding text is reworded. No attribution is given. How would you feel about that?

License fragmentation is bad

License fragmentation breaks the remix culture that Creative Commons seeks to establish. With this proposal, we would have

• Pre-2016 contributions, which are CC-BY-SA
• Post-2016 code under MIT-but-no-attribution-required
• Post-2016 code for which the poster elected to use the standard MIT license
• Post-2016 non-code contributions

Distinguishing what content falls under which license seems like a bit of a nightmare to me. And that's not even considering edits on top of existing posts, or questions that are loosely inspired by other posts (as often happens on Code Review).

In summary, please don't do this to us.

Answer 2

Please don't add a blanket exception to the attribution requirements.

I live on the Open Source site. Now, if someone came along and asked a question along the lines of "is it ever a good idea to add an exception to the requirements of an existing license" (and, lo and behold, someone has now asked that question), I'd tell them it's not - for these reasons:

1. You're essentially creating a crayon license.
   If you modify the terms of an existing license, you create what is known as a crayon license. Those are a problem - see "How can a “crayon” license be a problem?" for the reasons why.
2. It's far too easy to claim an illegal use is legal.
   See this answer of mine. Essentially, someone can get my code from somewhere I use it - perhaps in a commercial open-source product of mine - and not attribute it to me. When I chase them down on that, they can simply claim "oh, I got it from Stack Overflow", and get away scot-free.
3. It's massively unclear how to reverse that requirement.
   Say I don't want to have this extra exception applied to any of my code. How do I note that? If I put a note in my user profile that all my code snippets must be attributed properly as per the terms of the full MIT, is that enough? Do I have to add a note to every answer? I'd rather not have to do that, but I also don't want to be chasing people down endlessly for misusing my code.

Make it easy to opt my code out of this exception, or don't do it.

Answer 3

What?!

Am I the only person that doesn't understand what this all means? This really needs a TL;DR.

I was under the impression that any answers I provided on Stack Overflow that contained code are to be used freely, otherwise why would we post the code?

What I want answered is What do I need to do as someone who wants my answers to be used freely? Nothing?

Answer 4

A few thoughts:

1. Don't pretend this is the MIT license

Don't pretend you're using the MIT license when you're not. If you need a unique license for a unique situation then just use a unique license. While crayon licenses in general are a problem, I think that this the Stack Exchange network is a situation where a new ultra-basic license would be warranted. And I know you'll use your legal team to prevent the problems with other crayon licenses.

2. It would be best to make a new license

If the purpose of this change is to make it easy for post authors and easy for those who would use the code in those posts, then what you're doing won't help. I think it would probably be best to write a new license which is very basic and allows reuse as long as a URL of the post is attached to the borrowed code. If you want to have any user choices, then it should be between this and the existing CC BY-SA. It would also be possible to give users the option to do a CC0 type opt-in too, but the default must be attribution.

3. The license needs to be self contained.

You can't have the license, and then list exceptions to it in the terms of service, and then also require users to check a poster's profile to see whether the exception hasn't been opted out of. If you want to have two options, then just have two completely distinct licenses. Users also need to be able to see easily what the license of each post is...

4. Each post should show its applicable licenses

Each post should indicate below it what the relevant licenses are. If the user has selected only CC BY-SA or CC0 then it could show it. Old posts would show only CC BY-SA (unless perhaps their author edits them and ticks a box saying "update to the new SE-Attribution license"). Posts edited by multiple authors would show only the compatible licenses: new posts without any user opt-ins would be the same as if they only had one author, but if the user did opt-in to CC BY-SA only, then the posts would be only CC BY-SA.

5. Fix your footer "attribution required" link

It is against the terms of the CC BY-SA license to require specific attribution formats. It is dishonest and disingenuous to keep that link in the footer.

Answer 5

I have two thoughts about this. The first is that the line between code and non-code is sometimes blurry. This means that depending on how we write answers, different things are protected differently. That seems odd. The second is that the change in license should probably be accompanied by a reminder about the community standards for plagiarism: just because the code is licensed permissively doesn't mean that it's OK to copy within Stack Overflow.

Where's the line between code and non-code?

I know that we discourage code-only answers, but there are cases when they're appropriate: namely, when the code is well commented and self-documenting. Some languages have documentation strings, comment block conventions, etc., to support this. Suppose I write an implementation of an algorithm for a Stack Overflow question. In order to make the implementation idiomatic, I include a nice description in a documentation string in the code. According to this change, the description is now MIT licensed, whereas exactly the same prose description would have been CC-BY-SA licensed if I had written it outside of the code block. This seems very, very, odd. What happens if I write some "code", but don't put it in a code block? Maybe I put it in the answer text, but in boldface because I don't need monospacing. Which license is it under? E.g., if someone asks how we can convert a number into a list of its digits, I can post this as a code only answer (that, I think, doesn't have the typical issues characteristic of code-only answers):

(defun digits (n &optional (base 10))
  "Returns a list of the digits of N when written in a specified the
specified BASE.  The digits can be extracted from least significant to
most significant by repeated division.  When N is divided by BASE, the
remainder is the least significant digit of N.  The quotient is then
taken as the new value of N, and the process is repeated until N
reaches 0.  Note that in the case that N is zero, DIGITS returns the
empty list, rather than a list containing 0."
  (do ((digits '() (list* r digits))
       (n n)
       (r 0))
      ((zerop n) digits)
    (multiple-value-setq (n r) (truncate n base))))

Now, if I posted it like this, I'd get attribution requirements for the explanation:

Such a function returns a list of the digits of N when written in a specified the specified base. The digits can be extracted from least significant to most significant by repeated division. When N is divided by base, the remainder is the least significant digit of N. The quotient is then taken as the new value of N, and the process is repeated until N reaches 0. Note that in the case that N is zero, the function returns the empty list, rather than a list containing 0.

(defun digits (n &optional (base 10))
  (do ((digits '() (list* r digits))
       (n n)
       (r 0))
      ((zerop n) digits)
    (multiple-value-setq (n r) (truncate n base))))

This seems odd, especially for languages with strong documentation traditions.

But one of the comments from a mod seems to indicate that even the "code block ≡ code" rule (which I just made up; I don't see any actual guidelines for determining what's code and what isn't in the announcement) isn't really fixed either (emphasis added):

Just to clarify, we never said code is anything inside a code block. It might be up to a user to decide what's code and what isn't, just like he/she may have to make a judgement about whether something crosses the threshold of originality. Ultimately, always, it's on a code user to decide how he/she wants to use what they find here.
– samthebrand♦ 12 hours ago

I think there must be some misunderstanding here, because it seems like this would make the official policy: "You posts are licensed under CC-BY-SA, except for the code portions, which are under a MIT license with the attribution requirement removed, and it's up to the discretion of the people who use your posts to determine which parts are code." That can't be the intent, can it?

I think this would get especially blurry with any questions about Markdown or HTML, where the answer "text" itself might also count as the code. E.g., if someone asks "How do I make text bold in Markdown?" to which someone replies "Like this (see source)." The same issues could come up with TeX/LaTeX, since some sites support MathJax.

That's obviously a contrived example, but it highlights that we need some clear guidelines about how to determine which parts of a post are code, and thereby licensed under the MIT license (and whether the author requests attribution), and which parts are non-code and thereby licensed under CC-BY-SA.

(One more edge case that's probably not really important comes to mind: if we do, as some users are suggesting, or at least asking about, decide that edits should no longer turn text into code blocks, or vice versa, what happens with all the posts that (mistakenly) use code blocks instead of blockquotes, or that use backticks instead of quotation marks?)

Remind users that attribution is still required within Stack Overflow

I had some concerns that the more permissive licensing of code would make it much easier for a badly-behaved user to attempt to justify code plagiarism. After all, if they can copy it without attribution outside of Stack Overflow, why not within Stack Overflow, too? Shog9 pointed out that this is actually covered in the Help Center, How to reference material written by others. You can look through the edit history for some earlier thoughts before that clarification.

Answer 6

Sorry but I think this is a pretty terrible idea.

Code on Stack Overflow, Server Fault, etc. isn't "code" in the typical sense of licensing where you're concerned about distributing it for use by others. It is an example of one way (but not the only way) in which a particular problem presented in the question can be solved.
It is more akin to code snippets in a textbook than a software project on github.

CC-BY-SA is (IMHO) the most appropriate license for code posted to a Stack Exchange site: We're licensing the documentation (including snippets/examples) to you. If you happen to cut-and-paste bits of it into your production environment that's fine - my understanding has always been that you should attribute it the same way you would if you posted it to another site (by a URL pointing to the source) and you're good to go.

We (contributors) are not providing any kind of warranty or statement of suitability for a particular purpose with our code snippets, or any other part of our answers. If you want to go ahead and make language to that effect prominent in the terms of service by all means do so, but I don't think muddying up the license terms for site content makes sense!

---

That said, if you feel we must have a separate license for "code" posted to the site taking an existing license with specific requirements and saying in your terms of service "…but it's OK to ignore requirements X, Y, and Z" is REALLY TERRIBLE -- you're creating two legal instruments (Terms of Service for the site and a License for Use/Distribution of the code posted therein) with conflicting requirements - this is the kind of stuff lawyers charge hundreds of dollars an hour to bloviate about in front of judges.

artofcode already spoke about the dangers of "Crayon Licenses", and I think those points are well made, so I won't reiterate their points here.
Suffice it to say I think you should either find a license without a notice preservation/license reproduction requirement, create one without that requirement (dear $_DEITY, another freakin' Open Source License! Kill me now!), or craft one specifically for Stack Exchange code snippets that describes precisely what attribution is necessary (e.g. a link to the page you got the code from) and what notice preservation is required, if any.

Answer 7

I'm not sure that this would be a good move for Code Review, as questions and answers there often involve substantial amounts of code that are intended to be public knowledge, but not necessarily freely incorporated into other projects' code bases — especially for private commercial gain.

I haven't thought through what the ideal compromise is, but I think that this non-trivial change should be reframed more as a question for discussion than a proposal for ratification.

Answer 8

Contributors agree to give code users permission to ignore the MIT License’s notice preservation requirement, as long as users give reasonable attribution upon request of the copyright holder

Sounds like a recipe for boilerplate

# I hereby explicitly request attribution under the MIT license
# for this code, as stipulated in Stack Overflow TOS § NN ¶ n.

popping up in half the code blocks on the site. I think we can all agree that this is noise.

It may also be legal noise; I have no idea if it counts, honestly, but I want attribution for my code. I also have no way to track who takes it from the site, so what else do you want me to do?

Answer 9

You don’t have to include the full MIT License in your code base. Contributors agree to give code users permission to ignore the MIT License’s notice preservation requirement, as long as users give reasonable attribution upon request of the copyright holder (or Stack Overflow on behalf of the contributor). This optional exception to the MIT License will live in our terms of service.

How will this be allowed/enforced/implemented on Stack Exchange itself? For example, if I add a disclaimer to every post I make saying I require it - will this be allowed?

What if I modify my user profile to include an indication of this? Is this equivalent?

If most regular users require attribution how is this an improvement to the status quo?

---

Non-code contributions will continue to be available for use under the terms of CC-BY-SA

Code contributions will be available for use under the terms of the MIT License

This might be pedantic but... what is considered "code" for this?

If I write an answer that talks about code, is it a mess of licenses? Where parts of sentences are CC-BY-SA and parts are MIT? What about pseudo code?

Answer 10

I have sent an inquiry directly to the OSI for a comment on this situation. Frankly, I am surprised that they would be recommending the changes that are proposed here. I hope to hear back from them shortly.

On the other hand, I would like to know more about what parts of this recommendation are suggested, or endorsed by the OSI? When you say "you consulted with them", does that mean "you spoke with them and they think it is a great idea?" Does it mean "We spoke to the OSI, they think it Sucks, but, hey, 'we consulted them'!"

Does the OSI have any statement regarding what they feel is relevant from a objective perspective?

Answer 11

Please make it clear that code will be dual-licensed, not just pure MIT.

At least, that's what I hope you're trying to do, but the way you've phrased it, it could be interpreted to mean that new code posted to SO after this ToS change will no longer be licensed under the CC-By-SA license, but only under the MIT license.

Such an interpretation could mean that people currently reproducing content from SO under CC-By-SA, e.g. on a programming blog, would now have to change their licensing terms if the content they used included any code posted after this change. The option to just make everything CC-By-SA and not worry about any other possible requirements might no longer exist.

Also, arguably, this might put SO itself in breach of CC-By-SA. The "viral" ShareAlike clause of CC-By-SA requires that, when any CC-By-SA content is used in a combined work (that is not merely a collection of separate, independent works), the whole work must be released under CC-By-SA. While SO as a whole, or even a single Q&A thread, might count as a mere collection of separate posts, it's hard to argue that a single SO post containing both prose and code would not count as a single work that must be released under CC-By-SA, if any of it is.

(The original author is, of course, allowed to release their own work under any license terms they want, however self-contradictory they might be. But SO, as a redistributor, must abide by the license terms granted to it — and if those terms for prose are still just CC-By-SA, then arguably they must include CC-By-SA for all content, for SO to be allowed to distribute it.)

^{Also, off-hand, I kind of wonder just how compatible CC-By-SA and the MIT license actually are. As a permissible license, I'd assume that the MIT license does allow redistribution under CC-By-SA, provided that the requirements of the MIT license — i.e. including, unless waived, the copyright statement, license text and disclaimer — are followed. What I'm not 100% sure about is whether or not those requirements themselves, if not waived explicitly or via dual-licensing, might not fall foul of some language in CC-By-SA that forbids distributors of derived works from imposing additional restrictions on them. Honestly, it's been a while since I took a close look at those licenses, so I'm not sure; I'm just pointing this out as one more possible reason to prefer explicit dual-licensing.}

Answer 12

GNU LibC didn't become MIT licensed just because I traced through it in a StackOverflow answer.

That you wouldn't in your right mind derive from that particular code in your own program isn't the point. It's that having a code excerpt put on StackOverflow is not an instant carte blanche on that code's license.

Others mentioned that regarding Code Review...someone wanting to get feedback on their GPL project may not want to see their code relicensed MIT just because they put up a page or two of a routine. If it was important enough to bother framing up and asking for community feedback about, it may be something they specifically don't want to see adapted and made closed-source somewhere proprietary.

Beyond that...an attribution line in the source and a link to the StackOverflow answer that provided it isn't just good for "following the license", it's good documentation. SE URLs to questions and answers can be shortened down to just the stem, so http://meta.stackexchange.com/q/271080 instead of http://meta.stackexchange.com/questions/271080/the-mit-license-clarity-on-using-code-on-stack-overflow-and-across-the-stack-e?cb=1. Someone can not only see the reasoning behind why code is the way it is, they can go check up at a later date and find new information or raise new critiques of whether that's the real answer.

I have two competing feelings overall:

• feeling one: If you're the sort of person who has to worry so much about licensing and lawsuits, and have to make sure there's not one bit of GPL'd code in your project (or whatever)...it should be your culture of development that pays the tax on fretting about it. Stallman was Right, FSF/EFF 4eva.

• feeling two: I don't want to answer emails asking if someone can use code I post here, because if I wrote it and posted it inside an answer on this site, I don't care and they can do whatever they want with it.

So I don't really know how that translates into an answer, other than to agree with those saying this is probably not something that should be done unilaterally and on a quick timescale. It may need a more adaptive solution, like a process of pinging a post for a license audit...then on an as-needed basis various factors are gone over.

(e.g. I might in my profile tick the "anything by me is ok for being MIT licensed", and then all that would have to happen would be someone would look and make sure I didn't source the material from elsewhere, and perhaps it could be 'vetted' for license even if I'm not on the site anymore.)

Answer 13

I haven't read all the answers here, so this is probably a repeat. But given that you're only giving us two weeks to react, I'll say my piece anyway.

This is far too rushed. You may have been working for a long time, but

1. Your proposal is clearly not ready yet. You don't have a clear picture of how it would work.
2. Two weeks (including what is a holiday period for a large part of your audience) is ridiculously short time for feedback.

PLEASE DO NOT CHANGE ANYTHING ON JANUARY 1ST 2016. Even February 1st seems far too close. I'm not fond of the status quo, I think we must do something about it, your proposal is certainly something, but it is not the right thing.

Your proposal is a huge change, since it would allow code posted on Stack Exchange to be used anywhere with no restriction whatsoever. Please do not call this “MIT license”. The MIT license is about as permissive as it gets, but it does require one thing: to maintain attribution. Removing the attribution requirement should more properly be called putting the code in the public domain.

Your proposal is legally dubious. Some jurisdictions, including the one I live in (France), grant the author of a work moral rights, which for the most part (especially in the case of software) are limited to the right to attribution. These moral rights cannot be waived by contract. Does the fact that Stack Exchange operate under US law mean that if you waive attribution requirements in the terms of service, this applies to French contributors to? I don't know — you'd better clarify that with your lawyers, otherwise this puts users of code posted on Stack Exchange by French contributors at risk.

Given the impact of the change, you really should first consult, and then inform, all the users on Stack Exchange. For consultation, meta is the right place — but not just this meta. In addition to MSE, you should consult sites where code is routinely posted, and whose denizens don't frequent MSE much. Keep in mind that Stack Overflow is not the whole Stack Exchange network. You should consult communities that have different code posting patterns such as Code Golf, Code Review, Mathematica, TeX & LaTeX, Unix & Linux, …. To inform users of a change of such magnitude, meta is not enough; it's a change of the terms of service, and must be communicated to users before it takes effect.

You have not managed to present the change in a non-contradictory way. Do not change anything until you can explain the change. You claim

we will be licensing all new code contributions to Stack Overflow and Stack Exchange under the MIT License.

But then you go on to say

Contributors agree to give code users permission to ignore the MIT License’s notice preservation requirement, as long as users give reasonable attribution upon request of the copyright holder

Give attribution upon request? That's… crazy. How would I know that someone is using my code, in order to make such a request? How would I contact them?

Oh, and by the way, since I don't know where this is heading, I hereby request that attribution be given in any creative work derived from my contributions on Stack Exchange. This includes both past and future contributions. This includes contributions in the form of executable code. This does not include contributions that do not meet the minimum threshold of originality to be considered copyrightable.

But to future-proof your work, we recommend you do one of these 2 things, or both:

A) Add a comment to your code that links back to the post where you found it, or
B) Comply with the MIT as it’s typically used, by including the full license text in your source

Wait: “to future-proof”? Proof against what? Against a contributor making an attribution request? That opens a huge maintenance cost. What on earth could possibly be the benefit of introducing such an uncertainty? Pick a license and stick to it. Make it a dual license if you think it's best, but don't make it something that can change later.

Sorry, you really need to go back to the drawing board.

Answer 14

I think that the change is in general a good idea. I'm not a lawyer, and I assume most people that use Stack Overflow and other SE sites about programming aren't either. Licensing can get very complicated quickly, and removing ambiguity and fuzziness is a good idea. The Creative Commons licenses weren't meant for code, and that leaves some issues with how to follow them when reusing code from any Stack Exchange site.

But I don't think this change goes far enough, there is still enough potential confusion left when copying code, or especially when you're not copying directly, but also rewriting it. This confusion might be mostly caused by my lack of knowledge in this area, but I guess most programmers aren't experts in copyright law either.

In my completely irrelevant and non-lawyer opinion, a very large part of the code on Stack Overflow doesn't meet the threshold of originality required to be copyrighted. A lot of snippets are short examples on how to use a particular library function or language feature. There aren't many different ways to use the same API, so I don't think it makes sense that this could be subject to copyright. This is pretty much based on my opinion on how the law should work, not how it actually works, so I might be entirely wrong about the actual legal issue.

If I would copy and paste a reasonably large piece of code directly, the rules are pretty clear, I have to attribute it. But what if I just copy it, and then play around with the code until I understand it? It might look pretty different afterwards, it arguably is a derived work and I should attribute it. What if the code is simply a short example showing how to use a particular function, do I really have to attribute that? But what if I read the code, look up the documentation of the involved feature, and then use that knowledge to write something that looks pretty similar, but I don't attribute because I didn't actually copy it? There's probably a good reason why people do clean-room implementations of copyrighted code, but once I've seen an answer on SO I can't unsee it.

I think some of my arguments in the previous paragraphs are rather unreasonable, but the problem is that just going by common sense in the area of licensing doesn't seem to be a good idea. The only way to remove all ambiguity is to use a license that grants unrestriced rights without any requirements, or putting the code into public domain. I don't know how well it is legally tested, but something like the Creative Commons CC0 license might work for this purpose.

It it still not entirely clear to me how this attribution exception is supposed to work. So I suggest to get rid of it entirely and use the least restrictive license you can find and avoid the issue entirely.

Answer 15

How does this work with edits?

Say I have an answer containing code, written in 2015. In 2016, I edit part of the code. Is the part I didn't edit still CC BY-SA? Or does my edit implicitly indicate it's all under this MIT crayon license? How is this indicated to readers unfamiliar with the site?

Answer 16

I'm primarily concerned by the lack of attribution requirements.

Right now, all content on Stack Exchange is CC-BY-SA, which requires attribution. However, the proposed change to the source code license won't require attribution. The original post does say that the users of code must give "reasonable attribution upon request of the copyright holder". I'd like to see more information about how this will work. Specifically, I want to ensure that whenever people use any content that I created that they obtain from Stack Overflow, there is a requirement for attribution. I'm not clear how that will happen.

I'm also not convinced that "a comment to your code that links back to the post where you found it" is sufficient attribution, especially in a closed source project where users can't see the source and therefore won't see that my original work was used as part of the project.

I do understand that copyright has restrictions. For example, some things cannot be copyrighted. I understand that if I can't copyright the code sample, then there's no need to license it and anyone is free to use it. Mad Scientist's post here talks about threshold of originality. But for things that can be, I do want to ensure that attribution is required uniformly across content.

Answer 17

Does this potentially make answering pre-February 2016 questions illegal after that date?

If someone asks on January 31st:

I have this code, it doesn't work, how can I make it work?

then two days later, I might know the answer and just copy his code, change a few characters to make it work and post that as an answer. Except, that I am not allowed to. His code is licensed CC-BY-SA but by putting it in my answer, I would license it as MIT. I am not allowed to do that.

Is this correct, or am I missing something?

Of course to make it legal I can answer it like

In line 5, you have to change the 3rd 'a' at position 16 to 'b' and in line x...

But if it is that cumbersome to answer, I might just not.

Answer 18

If this happens, I think it needs to be very explicitly shown on each code block what license it falls under. Having multiple licenses co-existing and trying to determine which one a particular code block falls under by looking at the date will get messy fast (especially if people have to manually add boilerplate text to require attribution).

Users should be able to set their preferred license in their profile and new code-blocks that they write will automatically default to this license, but can be overridden if needed.

Old code blocks will default to the existing license but users can go back and edit to change the license going forward (and maybe provide a mass-update tool if people want to change the license for all their answers at once).

Answer 19

I don't get it.

Sorry, I fear this is exactly what you wanted to prevent, but I don't understand how to attribute any more.

You don’t have to include the full MIT License in your code base. Contributors agree to give code users permission to ignore the MIT License’s notice preservation requirement, as long as users give reasonable attribution upon request of the copyright holder (or Stack Exchange on behalf of the contributor). This optional exception to the MIT License will live in our terms of service.

WTH does this mean? By shortly skimming over it, this sounds like I would not need to give any reasonable attribution whatsoever, until the contributor asks me to do so. This can't be right.
After reading it again, the "ignore" seems to refer only to the "License’s notice preservation", which is… wait… only the thing that follows the Copyright (c) … line. Right?

I fear that this is what will happen for most people. They read "You don’t have to include the <License stuff>…" and think "Ah good, no attribution required". This should not happen. So please be explicit in your licensing requirements, and state something like

You have to include the line

Copyright (c) 2015 [Bergi](https://stackoverflow.com/users/1048572/bergi) at http://meta.stackexchange.com/a/271299/183280, MIT-licensed

above the used code, and should (if possible) or must (if requested by the copyright holder) include the full terms of the MIT License below that.

Answer 20

always been a little ambiguous how CC-BY-SA covers code

No, most ambiguity is gone at this point. CC BY-SA 4.0 now one-way compatible with GPLv3

You are changing from a copyleft to a permissive license. A lot of people, including me, care that our contribution is "shared alike," (the sa in cc-by-sa).

https://opensource.com/law/13/1/which-open-source-software-license-should-i-use: "the choice of license reflects as much of the social contract that is being made for the project as any legal requirements" The the social contract should not be changed unilaterally like this.

Edit: after reflection, I'm mostly ok with the change because copyleft for small code examples is usually more effort than it's worth (gnu recommends against copyleft for small < 300 lines https://www.gnu.org/licenses/license-recommendations.en.html), but I still do not agree with how it was changed, and think the reasons I mentioned are still valid reasons for being against it.

Answer 21

I agree that requiring the inclusion of the full MIT license for a small snippet of code from SO is unreasonable, but I feel entirely waiving attribution is also unreasonable, and directly against the principles of user-contributed content SO was founded on.

I propose that attribution continue to require a link to the SO post where the code originated and the name of the poster, in the same way that CC-BY-SA attribution works now. In open-source code the standard "link in a comment" practice you recommend should suffice. Commercial code will add the link in the usual "disclaimers and licenses from third-party code" spot; they might choose to include the full license anyway to keep things simple for their lawyers.

(I think this is basically using CC-BY for code instead of CC-BY-SA for prose, but I am not a lawyer.)

Answer 22

Here's my two main points:

How is it shown what license a post is licensed under?

Once posts become licensed under different licenses, how will we know which license is used by a particular post?

• Will there be an icon next to your flair at the bottom of the post which indicates the license?
• Will code blocks be colour coded based on license?
• Will a footer be added to every code block stating the license?

Whatever approach is used, it needs to be clear and understandable to everyone, even those who haven't used SO before.

How can I make it clear that I still want my code to be attributed?

Contributors agree to give code users permission to ignore the MIT License’s notice preservation requirement, as long as users give reasonable attribution upon request of the copyright holder (emphasis mine)

I may request to have all my code attributed. How can I make this clear to those who use my code?

@Josh Caswell stated that you can probably use something along these lines in your code blocks:

// I hereby explicitly request attribution under the MIT license
// for this code, as stipulated in Stack Overflow TOS § NN ¶ n.

Is there a way this can be implemented into the UI (i.e a checkbox which adds a notice to the bottom of the post or a preference in your settings which automatically does so to all posts) or will we just have to do this ourselves?

Answer 23

• If I'm reading you correctly, henceforth code is available only under the MIT license (or your modified version thereof), not the CC BY-SA license. Why not retain the existing license, and add the MIT license? This would make life a lot easier for people copying posts from here. It's much easier to copy an article if the whole thing is under one license.

• Define, please, "code".

  This condition is met when a > 3.

  Is that "code"? Is most of that sentence CC BY-SA, but those final characters MIT? Is this not a mess?

• If I edit someone's post to format their code properly, am I changing their license terms? Do I have legal permission to do that?

Answer 24

If people are using code I've written on SO, but are not publishing the MIT license with attribution... What's to stop me or someone else from suing them proclaiming they didn't receive a copy under the MIT licensed form published on SO? They are using unlicensed versions of my code at that point...

I have a hard time believing I won't have a leg to stand on in court exerting my IP rights when they've no indication of licensure for said code of mine.

The MIT license I extend to SO is given to them but I still maintain personal IP rights to productize said code and sell it to others holding them to separate licenses than that which SO and people who receive the code on SO are given. Per SO's TOS I am still the right-holder, they are merely provided a license by me upon my submission.

Honestly, I think the whole MIT licensure thing is a perfectly good idea. I think that "you don't need to publish the license upon your usage of the code" exception cannot function reasonably, and if utilized opens loopholes in the whole concept of licensing the code to begin with.

Answer 25

Contributors agree to give code users permission to ignore the MIT License’s notice preservation requirement, as long as users give reasonable attribution upon request of the copyright holder (or Stack Exchange on behalf of the contributor).

There are two fairly obvious problems with this:

1. It's not always possible to add attribution later when requested by the copyright holder. E.g. if someone makes a paper book which teaches a subject using code taken from various posts, and then the authors of some of those posts request attribution, the best that the publisher can do is to pulp the unsold copies and put an erratum on their website.

2. About half of the MIT licence is a disclaimer of liability. Replacing the notice preservation requirement with "attribution" means not needing to state that the author and copyright holder has disclaimed liability. If you're saying, in effect, that half or more of the licence you've chosen is irrelevant and unnecessary, why on Earth did you choose that licence?

Answer 26

If you don't want people to use your code snippets, why did you post them to Stack Overflow?

Reality: The purpose of the CC-BY-SA licence is to stop link-farmers ripping off the content to auto-generate a vast Internet of Crap. It's not to make sure you get a mention in the comments in Contoso's internal web portal because you knew the correct way to deserialize a widget.

The new licence should be that the code snippets in answers are public domain, or at any rate are licensed with the only obligation not to hold the author liable, and no other obligations.

If you have some code you want to protect with copyright, it's not appropriate to post it to Stack Overflow.

Answer 27

I think there's been an excessive knee-jerk reaction against this change. I'd encourage the skeptics to:

Read the label

Stack Overflow is a question and answer site for professional and enthusiast programmers.

Source: https://stackoverflow.com/tour

The mission-statement for Stack Overflow, at least, is to provide a useful resource for both professional and amateur/enthusiast programmers. It's intent is not to primarily benefit the open-source community (or any other niche), nor to take sides in an ideological debate on IP/software rights and licensing.

The 'professional' part is important here, as it implies operating for commercial gain. Typically in the employ of a corporate entity, and/or with at least some codebases that are not open-source or otherwise publicly available.

Responsible professionals know better than to incorporate copyleft code into a proprietary/closed-source codebase. Which makes CC-BY-SA licensing a very big problem (technically speaking), to the point where it's in direct conflict with SO's published mission-statement.

 

Understand CC-BY-SA

CC-BY-SA is essentially a copyleft license. Note that the 'SA' in CC-BY-SA stands for "ShareAlike", which is a licensing requirement that any derivative works which incorporate CC-BY-SA content also be released under CC-BY-SA. And that's where the copyleft problem comes in for professional use.

Interpreted in the context of a software codebase (though as noted in the OP, CC-BY-SA was not really intended for use with software in the first place), it means that if you use a CC-BY-SA licensed snippet in a codebase then strictly speaking you must release the entire codebase under the same license should you distribute any works derived from your codebase (such as a complete application/compiled binary).

Or in other words, copy a snippet from SO into your Android application, then by a strict interpretation of CC-BY-SA (or any other copyleft license) you must release your application's source-code under the same terms if/when you submit it to Google Play. That makes code licensed under CC-BY-SA technically unusable for most professional (i.e. commercial) purposes.

Which, again is at odds with SO's stated purpose.

 

Consider the reality of the situation

The SO community advertises itself as a place for professionals to collaborate and discuss solutions to technical challenges. And there are undoubtedly plenty of professionals to be found here. And certainly some of them have probably used some code snippets in proprietary, commercial applications. And it's certainly likely that an even greater number of non-members who find a useful snippet on SO through a Google search have used it in closed-source projects of their own.

To pretend that this hasn't happened, doesn't happen, or isn't happening right now is silly. To argue that it shouldn't happen is a denial of SO's published mission-statement. And to try and force the issue and litigate someone over violating the CC-BY-SA license on a code snippet would be virtually impossible if they don't self-nominate (how on Earth are you going to find out that they used your snippet?).

To have all of the code on SO licensed under the terms of CC-BY-SA is actually a massive impediment to the site's intended use-case. It's incongruous with the way that people actually use the site, in practice. On paper, it's a massive risk for any professional users. And in actual fact, it's probably an issue largely ignored by the majority of users (and non-users) who turn to SO for coding help.

The only thing the new changes do is bring the 'on paper' licensing terms in line with the way people actually use SO. Which is something that we should all welcome.

 

Consider why you post code on SO

Are you here to try to help people solve problems? Or are you here to get your name listed against as many projects as possible, one small snippet at a time? Or to play 'gotcha' with unsuspecting professionals, litigating them when you catch them actually using the code you gave to them?

Unless you're here for the third reason, you don't need to have your code snippets licensed under copyleft! And if you are here for the third reason, you're participating in a highly ideological debate. Which is entirely your right, but which is also something that SE probably shouldn't get entangled with.

And besides, would you really litigate someone for using your code snippet commercially, and do you believe it would really be ethical to do so? Considering that you know there's a good chance the person asking the question is a professional developer, and that you know that professional developers often work on commercial, closed-source code. If you object to your code snippet being used in such a context, then why are you answering their question with code to begin with?

If your motivation is to help others, the licensing changes do not impact you in any way. Or, they might make it possible for you to help even more people, as the new license makes your code usable in places where CC-BY-SA code technically is not.

And note that there's really no issue with respect to attribution. CC-BY-SA requires attribution, and the new license terms still permit you to require attribution. So you can have your attribution if that's what you're after.

The substance of the change is removal of the 'ShareAlike' component from the license, which changes code on SO from being copylefted to being permissively licensed. It's very similar to changing the license terms to CC-BY, which requires attribution, but drops the 'ShareAlike' requirement.

 

Consider what's fair and reasonable

Forget for a second about the legal precedence that establishes that IP creators can essentially dictate terms according to whatever arbitrary criteria they choose.

Instead look at how much time and effort actually goes into code snippets posted on SO. FGITW means that the average snippet is authored in minutes, not hours. A good FGITW player will at least invest some additional time curating and improving their snippet after it's posted. However, all up it's probably very safe to say that most code snippets on SO are authored in 60 minutes or less.

So what's a fair reward for 30-60 minutes (at most) of effort? Is it a thank-you, some +rep, and being able to (should you so desire) require attribution from people who find your code helpful and use it? I'd say that's fair enough. You made a helpful contribution, and are entitled to recognition for the effort.

But is it fair or reasonable to demand that the person who uses your snippet release their entire codebase under CC-BY-SA, when that codebase may contain many thousands of lines of code and be the product of many person-years of combined effort? I think not. The snippet is a drop in an ocean; it shouldn't allow you to lay claim to the entire ocean.

But that's exactly what you could do if code snippets are licensed under CC-BY-SA (assuming no legal hurdles and technicalities get in the way). And personally I think that's absurd. The value of an answer is real and nontrivial. But it's not so large that providing someone with a code snippet should mean you have a claim to every codebase that snippet gets used in. Attempting to do go goes against the spirit of collaboration, community, and helpfulness that underpins SO, in my opinion.

As such, the new license terms strike a better balance on the 'fair and reasonable' scale, and address a potential imbalance between the value provided by a CC-BY-SA code snippet and the scope of the claim that can be staked against a much larger codebase by leveraging the 'ShareAlike' component of CC-BY-SA.

 

In closing

I welcome and fully support this change (though I think the suggestions to implement this as a dual-license as opposed to a wholesale swap have merit, so long as it's the consumer of the code who chooses which license they want). You should too. There's no reason not to. Information wants to be free.

The sooner this can take effect, the better.

Answer 28

Reasoning

I don’t think this is a good idea.

Background

I’m a BSD developer myself, so I’m normally happy with putting things under BSD/MIT-style licences, but there are two big things that go against BSD spirit there:

The exception from reproducing the licence ☹

This is really bad, as attribution is all a BSD developer usually gets (and, occasionally, patches to the eMail INBOX). It also creates a “net” licence by itself, the problems with which others have already pointed out.

So, if you’re going to switch to MIT, use MIT, not MIT with some exception.

What is code?

I’m a very big fan of putting everything under the same licence, be it code, data, documentation, etc. so I cannot help but disagree with distinguishing between “Non-code contributions” and “Code contributions”. In addition, someone has to decide what is what, and that usually falls to judges, and we all know things just go down the drain when lawyers are involved.

This means:

All three bullet-points you stated are against the spirit of those who use MIT-style licences in the wild the most.

Other problems

Some people have pointed out that e.g. “the GNU libc doesn’t become MIT because I traced through it on SO”. This is true, you obviously can only licence content where you have the permission to do so. However, including such information on SO posts is still necessary; this is just as true currently though, so there needs to be an amendmend to the ToS that it’s fine to post content one does not own as long as it’s under an OSS licence. As things currently stand, HostileFork admitted what’s basically violation of the ToS by posting content where he doesn’t have permission to post it under CC-Wiki.

My own solution

I’m okay with the current situation, where the entirety of the post is published (harmonically) under a weak copyleft licence, but where I have the following snippet on my network profile:

I hereby offer all content I author on the SO/StackExchange network also under The MirOS Licence (HTML version with less WTF-8) in addition to “cc-wiki” (CC-BY-SA 2.5 or 3.0, depending on the SO/SE network site, apparently). I urge everyone else contributing here to do the same.

(I should probably plug the full terms there; had to do so for opencaching.de (on which I put my geocache listings under those terms as well) have requested I do that and add a German translation as that site is also used by many people who don’t speak English… which I did.)

This allows authors to choose more free licence terms for their posts (of course I’d be able to exclude individual postings, and postings in which I posted content that’s not mine, such as GNU libc excerpts), but preserves the copyleft-and-attribution-requirement of the general site.

And everyone who does take content of mine under The MirOS Licence will note it also has a requirement to retain full attribution and licence terms (all three of copyright statement, terms and discaimer), while permitting inclusion in other works.

Circle back to harmony of code, data, documentation, media content, etc: this licence specifically uses the term “work” instead of “code” while otherwise being a mostly standard BSD/MIT/HPND-style licence. This is precisely to facilitate putting everything under the same licence terms.

What SO/SE should do, IMHO

• Stop this hurried action
• Add some wording to the ToS about excerpts from works one does not own (permit it as long as the work’s under some OSS licence), as that’s as much a problem with the current terms as with the new proposed ones
• Rethink, read community feedback
• Possibly allow opt-in or even opt-out (per (site, not network) account) to use a permissive licence but do not waive the notice preservation requirement for them
• Allow opt-out of the permissive licencing for individual posts

This makes it fine for sites such as Code Review to continue operating on the current terms (it’d probably qualify for opt-in) while still permitting you to put e.g. SO under an opt-out scheme for more permissive licencing terms (with more advance notification, this time).

I do hope any changes to the terms such as these would be prominently featured in a to-be-acknowledged page the first time anyone posts after they go live, right?

Answer 29

Well... I see after only three hours many people have some very strong opinions about this.

I opened a question about clarifying SO's intentions on licensing a while back so I would simply like to say thank you.

Thank you for taking the time and resources to address this issue. I'm sure there were a lot of tough decisions to be made and I'm sure handling the reaction is going to be pretty difficult as well.

So again, thank you for addressing this.

To the people who want to offer an opinion on the outcome I'd recommend you read this question and others before throwing in an opinion. It breaks down the complications with the prior license and seems to demonstrate interpretations fell one of two ways. If you want to have a constructive impact on what's happening it will benefit to be aware of what has already been discussed.

Answer 30

So we've got

1. Contributors who want their code to be used as freely as possible,
2. Contributors who want their code to be used only with attribution,
3. Programmers who want to learn from code examples without copying from the code itself,
4. Programmers who want to use code under clearly defined terms so they can steer clear of copyright infringement, and
5. Programmers who just want to use code in their projects without having to worry about all that legal mumbo-jumbo.

Keeping everything CC-BY-SA works well for group 2. Maybe for group 4 as well, except that CC licenses (aside from CC0) are generally not recommended for use on software.

MIT-licensing all code from 2016 onwards is also fine for group 2. It might help group 4, except you don't know for sure whether the user who posted the code is actually the author... or got it from somewhere outside StackOverflow... or modified it from another StackOverflow question or answer which isn't necessarily MIT-licensed... etc.

A quasi-MIT license that makes attribution optional, unless the author comes and asks for it later, adds future uncertainty and thus is even worse for group 4.

A less restrictive option would be to make code contributions available under the Unlicense, the CC0 Public Domain Dedication, or similar "do what you want" terms. That would satisfy groups 1 and 5, although group 2 wouldn't approve. Group 4 might be a little worried as well, since there's some dispute over whether waiving copyright is legally valid (though that's a much longer discussion).

Group 3 would be satisfied with pretty much any licensing terms.

---

I think the best way to strike a balance is to make everything CC-BY-SA by default, and allow individual contributors to share their code under more permissive terms if they so choose.

A good start would be providing an option to add licensing/permission notices in user profiles. For example:

"All the code I contribute to StackOverflow is copyright <My Real Name> and open-source under the MIT License (http://opensource.org/licenses/MIT)."

Or

"All the code I contribute to StackOverflow is available under the CC0 Public Domain Dedication (https://creativecommons.org/publicdomain/zero/1.0/deed.en)."

This helps groups 1, 4, and 5 by lifting at least some restrictions on code sharing. It also satisfies group 2, since the attribution requirement is preserved for those who want it.

---

I won't pretend there's a perfect solution. Certainly not for groups 4 and 5. The safest option for avoiding copyright infringement is always to write your own code -- possibly using the same ideas you've seen elsewhere (since ideas can't be copyrighted), but expressing those ideas in your own words.

And from my perspective, the central purpose of StackOverflow is to learn. Not just to learn how to fix some specific code, but to gain the ability to solve problems of that type in any form. And for that purpose, reading the answers to learn from them -- without copying their text -- is perfectly fine.