Why are you re-licensing ggplot2?

[slowkow]

@hadley said:

We are systematically re-licensing tidyverse and r-lib packages to use the MIT license, to make our package licenses as clear and permissive as possible. To do so, we need the approval of all copyright holders, which I have found by reviewing contributions from all all non-RStudio contributors.

#4232 (comment)

I asked:

Could I please ask if you might provide some context about the reason for the change of license?

I'm afraid my comment was not visible, so I'm posting a new issue here, where I hope ggplot2 users and contributors might learn more about the plan to re-license. I would like to please ask if someone might be willing to provide some context regarding the reason for the change of license. Is that OK?

ggplot2 is available under the terms of the Free Software Foundation’s GNU General Public License (GPL). https://github.com/tidyverse/ggplot2/blob/master/LICENSE

That's the same license that the R Project uses:

R is available as Free Software under the terms of the Free Software Foundation’s GNU General Public License in source code form.

For those of you who are not familiar with software licenses, you might consider reading this recent post from a company that re-licensed their code from from the MIT to a newer licensing scheme called GNU Affero General Public License V3 (AGPLv3).

For example, one reason to re-license code from GPL to MIT might be to give permission for a corporation that is interested to take the ggplot2 (and other tidyverse and r-lib) code and use it in closed-source proprietary products.

[clauswilke]

Here's my perspective. But first, I'd like to emphasize that I had nothing to do with this decision, and didn't even know until the issues were filed. I also have absolutely no financial interest in any company that could benefit from relicensing.

In theory, the GPL protects you from cheaters that use the free code without contributing back to the community. In practice, this doesn't work, and the people actually hurt by the GPL are other free software developers that have good intention but that for some reason or another need to work in a licensing environment that is incompatible with the GPL.

The problem with the idea that the GPL protects you is that companies that are willing to cheat the system aren't particularly disturbed by violating the GPL. There are plenty of high profile examples, most recently Tesla violating the licensing requirements of the Linux kernel. And these cases never go anywhere. You can sue the company, and it costs you a lot of money, and usually the best-case outcome is that the company just posts a current dump of their code base somewhere, so they're in compliance temporarily, and then they go right at it again violating the license.

The only people that actually take the GPL seriously are other free software developers. And the most common problem they run into is GPL incompatibility with other licenses. The GPL is even incompatible with itself! (Version 2 vs version 3.) So, if you're developing some R package licensed under GPL 3, are you free to incorporate some ggplot2 code? Probably not. So, are you even allowed to write any ggplot2 extension package that is not GPL2 licensed? I'd argue that any ggplot2 extension package is a ggplot2 derivative, because you basically can't write a new geom, stat, theme, or coord without first copying some ggplot2 code. So, to arbitrarily pick one example, the ggdist package is in my opinion entirely out of compliance, because it is GPL3 licensed and obviously a derivative of ggplot2. However, I also think this is a silly notion, and we should allow extension package developers to license their code as GPL3, or MIT, or whatever they choose. So, the ggplot2 license needs to be more permissive. MIT is a good choice. I've recently used MIT for all my packages that are not clearly ggplot2 derivatives, and I've grudgingly chosen GPL2 for packages such as ggtext that are derivative in nature. (Compare to the license of gridtext. The two packages are companion packages and were developed in parallel.)

For these reasons, I strongly support the relicensing effort.

[hadley]

We're working on a blog post that explain this is more detail, but it's pretty much what it says on the can — the tidyverse currently uses a mix of licenses (GPL-2, GPL-3, GPL >= 2, MIT) and that causes needless confusion. Since it's quite a lot of work to relicense all the packages, we only want to do it once, so we opted for the most permissive license to enable the widest possible usage. While this licenses certainly allows you to make a proprietary fork, and this is something that I worried about when I first created ggplot2 (hence the GPL-2 license), I don't think it's possible to build a business around it — it's very hard to sell something that's 95% the same as a product that you can get completely for free.

@clauswilke Personally, I don't believe that extending ggplot2 in that way forms a derivative work — I don't think merely using ggplot2 or building on its exported extension interface implies that you need to be compatible with its license. (This is the same reason why an R package can have any license, not just GPL). That said, a lawyer might argue differently, and we do want to make it as clear as possible that you're free to build on ggplot2 in any way that you want.

[clauswilke]

@hadley To be clear, I completely agree that the licensing provisions of the GPL end at package boundaries. Since R is not compiled, the code only gets combined at the enduser level, and the GPL doesn't speak to that. It only cares about distribution. I can distribute my non-GPL package using your GPL package without ever distributing your package myself.

What is less clear, though, is the situation where people take ggplot2 code and modify it for their own purposes. E.g., I copy the code of geom_point() and then I modify it until I have a geom that can make a scatter plot of kittens. I think that's derivative code. I also think most ggplot2 extension packages are developed this way. I know all of my own started this way.

In any case, I don't actually want to worry about these kinds of things, which is why I very much support less restrictive licensing agreements.

[slowkow]

Thanks for the quick replies. I look forward to the blog post that explains this in more detail.

@hadley Here's one question you might consider addressing in the blog post:

• Why do some projects like RStudio (AGPL v3) and R (GPL v2) continue to use a GPL license, while other projects like ggplot2 are re-licensing to MIT? What are the main reasons why should users and contributors should favor re-licensing for one project but not the other?

@clauswilke Could I please ask if you might elaborate on this comment?

the people actually hurt by the GPL are other free software developers that have good intention but that for some reason or another need to work in a licensing environment that is incompatible with the GPL.

Could you offer a specific example that might be instructive for people who are looking for jobs and might not know the consequences of joining a company?

[clauswilke]

@slowkow I think making it about joining a company versus not is missing the point. The situation can be as simple as two open source projects not being able to exchange code snippets due to incompatible licenses. Here is an example of a case I've been involved with, which is somewhere between those two extremes: clauswilke/PeptideBuilder#9 (comment)

[hadley]

@clauswilke along those lines, I feel like I have recently benefited from the permissive licensing of a few other projects (particularly in the js community) and it’s nice to pay it forward.

@slowkow we can only really speak to why we made the decision to change. It is a lot of work (particularly if you’re not on GitHub) so it’s not surprising that most projects stay from the beginning. Also note that the RStudio IDE uses a CLA so it can be relicensed easily. We chose not to do that for the tidyverse to make it clear that the packages are community “owned”.

[has2k1]

The only people that actually take the GPL seriously are other free software developers. And the most common problem they run into is GPL incompatibility with other licenses.

I have had this issue with Plotnine which qualifies as a derivative of ggplot2 and so had to be licensed under GPLv2. The Python ecosystem in which it exists is dominated by BSD 3-Clause and MIT licences and it has made some anxious about working on top of it, or simply using it in a commercial environment.

Also, as GPLv2 is quite involved ("lawerly") it is less precisely understood; this has some users who care not to violate licenses to sometimes incorrectly assume incompatibility. So re-licensing is mighty welcome.

[yutannihilation]

I'd also like to emphasize the problem of incompatibility. I've seen this became a problem several times e.g. r-lib/covr#256 (comment). Actually, I myself had experienced a case where I needed to reinvent stringr functions just because its current license (GPLv2) was incompatible with GPLv3 packages. I don't think it needs to be cared, but it's not easy to convince the companies that have a stricter view on licensing issues (and I understand they try to avoid the potential risks even if it's very small). Migrating to permissive license let us avoid such unnecessary confusions (at least to some extent), so I hope relicensing will success.

[slowkow]

I'm still looking forward to the blog post, and hope that there will be some discussion of costs and benefits.

It goes without saying that every decision has costs and benefits. However, enumerating and weighing costs and benefits might help to compare the the ratio of their weights and to eventually arrive at a rational decision — or at least to reveal the beliefs and values of the decision maker.

---

Here, I want to offer some questions for your consideration.

Several people have mentioned incompatibility as a reason for changing from GPLv2 to MIT. You've described it as "unnecessary confusions", "lawerly", something you "don't want to worry" about, and "needless confusion."

• Is the confusion experienced by people who are learning about software licenses a cost that should be minimized?
• Is it possible that the cost of learning about software licenses is balanced by the benefit of understanding them?
• Alternatively, is it true that there is no benefit for learning about software licenses and user freedoms?

Incompatibility may come as an unwelcome surprise to those who use GPL before understanding it.

Proprietary software developers have the advantage of money; free software developers need to make advantages for each other. Using the ordinary GPL for a library gives free software developers an advantage over proprietary developers: a library that they can use, while proprietary developers cannot use it.

https://www.gnu.org/philosophy/free-software-even-more-important.html

We might consider whether the cost of incompatibility is worthwhile in exchange for the benefits of GPL.

• Are we interested to use a software license to protect users' freedoms? Do we weigh this as something important, or is the cost of this protection too high and the benefit of protection too low?
• Does the benefit of freely giving code away to other open-source (but non-GPL) projects outweigh the benefit of using GPL to protect users' freedoms?

[brodieG]

I think we should be careful dismissing the GPL on the basis of some "cheaters will cheat anyway" argument. There are many counter-examples.

I really dislike that for profit entities can take MIT code and use that for profit without any requirement to give back to the OSS community. Even the possibility of a lawsuit is enough to modify behavior, however hypothetical it might be.

Yes, GPL is more restrictive, and a little more annoying. For the GPL2 vs 3 compatibility issues, the code can be licensed under both.

[hadley]

@brodieG practically, I think the major impact of threat of a lawsuit is to push companies towards permissively licensed products. But regardless of what you believe the impact is, I hope that you agree that it's up to the primary developers of the individual package to choose the license that they feel is most appropriate.

[brodieG]

Which is why it is a shame when GPL licensed software moves to MIT, thus making it easier for companies to avoid the requirements of GPL. And yes, in general I agree copyright holders are entitled to license however they feel appropriate under the constraints of whatever software licenses they themselves are using. Although again, a shame that it makes it easier for for-profit entities to take OSS code without giving back.

As an aside, I personally think (though there is definitely some gray here), that R packages used with GNU R should be GPL-2|3. Of course, it is possible to use R packages with other S/R implementations that are not GPL that may not have such requirements. I realize not everyone (including R-core) agrees with this interpretation.

[yutannihilation]

I hope that you agree that it's up to the primary developers of the individual package to choose the license that they feel is most appropriate.

To be fair, if being GPL was one of the reason that a contributor decided to contribute, they has right to disagree. There might be such kind of contribution that ggplot2 couldn't have received if it started with a different license.

On the other hand, I guess there are some developers who gave up ggplot2 because of its being GPL. For example, as @has2k1 pointed, porting to another language is a derivative work, and probably it might not be easy depending on the type of the language or the ecosystem. This means ggplot2 might lost some possible contributions e.g. feedback about nicer API that we won't be able to come up with. I now feel it's wrong to call this as "needless confusion."

(Btw, here's one more example I happened to find. If dplyr kept uisng GPL, Apache Arrow might not have support for dplyr: https://twitter.com/wesmckinn/status/393122387610198016.)

[slowkow]

For those of you looking to learn more about the reasoning that goes into re-licensing, consider reading some of the following articles.

After reading them, I've become aware of a possible cost that might be worth mentioning:

• After re-licensing ggplot2 to MIT, ggplot2 won't be able to accept code copied from GPL projects. This might sometimes be resolved by asking the donor project authors to re-license from GPL to MIT.

• What happens to the code that might have been copied from a GPL project to ggplot2, now that ggplot2 is re-licensing to MIT? I don't know.

I'm not aware of compelling reasons for favoring MIT versus BSD.

In favor of BSD:

The Whys and Hows of Licensing Scientific Code

There is a tradeoff to choosing a BSD license, of course. It means that you’re effectively putting off-limits the incorporation of existing GPL-licensed functionality in your package. Hunter mentions some specific challenges this has posed for matplotlib; I’ve seen similar challenges within packages I’ve worked on, namely SciPy and Scikit-learn.

Why we should be using BSD

There are two unpalatable options. 1) Go with GPL and lose the mind-share of the private sector 2) Forgo GPLd code and retain the contribution of the private sector. This is a very tough decision because their is a lot of very high quality software that is GPLd and we need to use it; they don’t call the license viral for nothing.

Relicensing yt from GPLv3 to BSD

When I initially decided on a license for yt (seven years ago) it seemed appropriate to use the licensing terms as a mechanism to encourage contributions upstream. However, within the current ecosystem, it is clear that because of the virality of the GPL and the prevailing mindsets of developers, it is actually an impediment to receiving contributions and receiving mindshare.

In favor of AGPL:

Open source licensing and why we're changing Plausible to the AGPL license

Here are a couple of events that made us aware of the risks with a permissive open source license:

• There’s been at least one case where a corporation has taken parts of our code, made it closed source and started selling it as a direct competitor

• We’ve had approaches from large corporations that want us to help them so they can sell Plausible Self-Hosted to their tens of thousands of clients without wanting to contribute anything to our project. They offer publicity in return