written on Tuesday, July 23, 2013
The GPL used to be the cornerstone of the Open Source movement. At least it always felt that way. Clearly if you looked closer the situation has always been a situation of many licenses and the GNU GPL was only a small part of it. In recent years however it has become painfully obvious that a lot of developers have built up an open hatred for the licenses on different levels.
What I found surprising however is how little the license is being discussed nowadays. If you Google for discussions on the GPLv3 the most you are generally greeted with links to articles from 2006 and earlier when the debate about the latest revision of the GNU GPL was at its height. However none of the concerns of the discussions back then were ever addressed and you can see today that this has left a big impact on how the GPL is seen today.
The whole topic got a lot of relevance for me again recently due to GitHub. GitHub is one of the pillars of Open Source development these days and yet it's also a place where there is more improperly than properly licensed software. Their attempt to change that has been a simple license selector and I believe that is a terrible idea, especially because it brings up the whole GPL topic again.
This article is opening up a bit of the history of Open Source software licensing, how it seems to change and what we could do to improve it.
Until the GNU GPLv3 released, the number one Copyleft license was the GNU GPLv2 . Copyleft and GNU GPL were seen as the same thing. GPL is a very restrictive license in the sense that it whitelists rights instead of just putting a handful of requirements into a license that otherwise permits everything. This made the whole topic of GPL compatibility was always something that people discussed. GPL compatibility is generally just the idea that a license can downgrade to the GPL. Most licenses were able to do that, but some licenses had some pieces of text that made this impossible. Most famously the Apache License 2.0 was seen as GPL incompatible due to extra restrictions on patents and some versions of the Mozilla Public License.
Sun liked the whole concept of not being compatible with GPL and later drafted a new license on basis of the Mozilla Public License version 1.1 which was deemed GPL incompatible and created the Common Development and Distribution License (CDDL). This allowed them to release software under an Open Source license without allowing the Linux developers to take advantage of their open contributions.
In 2007 the whole topic of GPL compatibility became a whole lot more complicated as a new version of the GPL was written. Due to how the GPL licenses work different versions of the GPL are incompatible with each other. That's not terribly surprising but it's important to note, because it has quite profound implications given how people license code under the GPL and how the ecosystem was supposed to work.
There is a lot of code out there that depending on how you look at it, it's either GPLv2 or GPLv3. This is due to it being possible to license code under GPL of a specific version or any later version. What defines what a later version of a GPL license looks like? The GPL itself :
If the Program specifies that a certain numbered version of the GNU General Public License “or any later version” applies to it, you have the option of following the terms and conditions either of that numbered version or of any later version published by the Free Software Foundation.
There are three camps currently: the camp that stuck to GPLv2, the camp that upgraded to GPLv3 and the camp that's GPLv2 or later and depending on the context is either GPLv2 or GPLv3.
The most vocal projects about their disagreement with the GPLv3 have certainly been Linux  and Busybox  which both decided the only applicable license is the version 2 of the GPL. On the other hand the vast majority of GNU code has been changed over to the GPLv3 a couple of years ago.
The fun effect of this now is that it's absolutely clear that there are two different worlds in which GNU and Linux are living. There is a certain irony that “GNU/Linux” is now a sort of license conflict. Since the vast majority of GNU projects are on GPLv3 only and Linux will always be GPLv2 there is no more code sharing that can happen between these projects as a result.
The GPLv3 switch of projects primarily shows up because these pieces of software stop being updated in Apple products or get removed entirely. The bash on my Mac is now seven years old because that was the last version released under GPLv2 and Apple does not allow GPLv3 software on their systems. It's not just my bash that has not been updated, most famously for developers GCC on OS X is stuck a few year in the past. This has also resulted in Apple backing clang and LLVM instead of the GNU compiler collection.
But why is Apple so opposed to newer versions of the GPL? The problem is that it has become clear in 2007 the GPL has become more of a political platform instead of a software license and started to put restrictions into the code that make it much harder to deal with. At one point the GPL primarily ensured that someone who distributes a piece of software under the GPL has to ensure that derived works are provided under compatible license terms. Now however the license does more than that.
The biggest issue for companies with the GPLv3 is most likely the “anti-tivoization” part of the license that has a whole extra section on how the license interacts when it's part of a consumer product. The exact definition of what a consumer product is a vague definition of “any tangible personal property which is normally used for personal, family or house hold purposes” or alternatively “anything designed or sold for incorporation into a dwelling”. It also says that in case of doubt a product should be seen as consumer product.
What happens if the product is a consumer product? In essence the license requires that a modified version of the software must be guaranteed to run on an unmodified device. The license requires that signing keys are in the open, that the manual must include information about how modified software can be installed and that the device has to ensure that modified software can generally be run. Thankfully the license does not require that the distributor is not allowed to void the warranty in that case.
Generally however the license terms are a huge problem for companies. For instance Apple sells devices with a secured bootloader that is a consumer product (all iPads and iPhones). It would be impossible for them to comply with the GPLv3 license requirements unless they completely get rid of the devices' security systems. And it's not just Apple. You won't be able to get your GPLv3 software in any app store, no matter how hard you try. The license restrictions of the AppStore, Google's Play Store and similar distribution systems and the GPLv3 are just not compatible.
|||the 451 group: On the continuing decline of the GPL|
|||GNU General Public License Version 3|
|||Linus Torvalds on the GPL Version being used (2007)|
|||LWN: Busy busy busybox|
That's however not the only thing that happened in 2007. Another license was published: The GNU Affero GPLv3 . In license text nearly equivalent to the GPLv3 it has some extra restrictions in that the license triggers in other ways than pure distribution. The license was written to make it possible to enforce the GPL in environments where no traditional distribution takes place. Primarily aim was taken at web services that use GPL software but only provide functionality through a web interface or API.
Since the GPL cannot be further restricted the solution to make GPLv3 and AGPLv3 compatible are provisions added to both licenses that all them to work together.
The AGPLv3 was a terrible success, especially among the startup community that found the perfect base license to make dual licensing with a commercial license feasible. MongoDB, RethinkDB, OpenERP, SugarCRM as well as WURFL all now utilize the AGPLv3 as a vehicle for dual commercial licensing. The AGPLv3 makes that generally easy to accomplish as the original copyright author has the rights to make a commercial license possible but nobody who receives the sourcecode itself through the APLv3 inherits that right.
I am not sure if that was the intended use of the license, but that's at least what it's definitely being used for now.
In parallel to all the new developments in the GPL environment, outside of it quite a few things developed. Not all of them had the same impact obviously, but they are countless and generally resulted in people looking at the GPL in a new light.
Toybox for instance is a project that's prime existence is to not be GPL licensed. It's being developed by Rob Landley who was the previous maintainer of the GPLv2 licensed Busybox project of similar scope. What's interesting about Rob Landley is that he's the person who pulled through the license clarification of Busybox to mean “GPLv2 only” instead of “GPLv2 or later” . Shortly after he did that, he left the project due to being annoyed with the efforts required to clarify the license and the license trolling shown by some contributors to the project .
Considering that Busybox is one of the projects that got infamous for enforcing the GPL through the legal system it's very interesting the former maintainer would start a new project under Toybox based on the BSD license. The change in thinking can be seen from some conversations on mailinglists and notes on his website. The most glaring one is most likely his statement about the effectiveness of the GPL cases:
From a purely pragmatic perspective: I spent over a year doing busybox license enforcement, and a dozen lawsuits later I'm still unaware of a SINGLE LINE OF CODE added to the busybox repository as a result... 
In the same blog post he mentions how the response to GPLv3 by Google and other companies has largely been boycott. In regards to why he now goes against the whole GPL and not just the GPLv3 is that Android and other projects are now trying to rid the whole system of the GPL. And that's something that can clearly be seen.
Android goes very far in providing a GPL free userspace. The general license information  generally tells people to license under the Apache License 2.0 with the exception of kernel modules which have to be GPLv2 licenses. Here again the whole irony of kernel and userspace being incompatibly kicks in. As to why Google likes the ASL:
We are sometimes asked why Apache Software License 2.0 is the preferred license for Android. For userspace (that is, non-kernel) software, we do in fact prefer ASL2.0 (and similar licenses like BSD, MIT, etc.) over other licenses such as LGPL.
Android is about freedom and choice. The purpose of Android is promote openness in the mobile world, but we don't believe it's possible to predict or dictate all the uses to which people will want to put our software. So, while we encourage everyone to make devices that are open and modifiable, we don't believe it is our place to force them to do so. Using LGPL libraries would often force them to do so.
Why are people so afraid of the GPL all the sudden? Partially because the GPL has always been a radical license. Especially in the absence of copyright reassignment. For instance the GPLv2 comes with a clause that has been dubbed the “GPLv2 death penalty” . Essentially it means that whoever violates the GPLv2 automatically gets the license terminated and not reestablished until they explicitly obtain a new license. Without one authoritative copyright holder it would essentially mean to ask each and every contributor for a new license.
Now in reality GPL violations have always silently reestablished the license for the violator once the dispute has been resolved, but the license does not actually state that. It's unlikely that this will ever matter as clearly enough court cases can act as an example of silently establishing the license again, but it does leave an ugly aftertaste.
More than anything it has become clear however that some think the FSF just cannot be trusted. There are two camps now around the FSF: the ones that believe into the general ideology pioneered by Richard Stallman and the ones that think the GPLv2 license is okay but that they are not okay with the direction the license is taking. Linus Torvalds obviously being a more prominent supporter of the latter camp. That camp exists because the Free Software Foundation is largely stuck in their own world  where cloud computing is the devil, cell phones are exclusively tracking devices and Android is something the GPL should prevent from happening. There are GPL supporters that don't support the current view of the Free Software Foundation which is dangerous, considering they are the only ones that are in the position to shape the future versions of the GPL. Even some GNU projects seem to be disagreeing with the goals of the GNU project and the Free Software Foundation. On December 10th 2012 GnuTLS split off the GNU project .
|||Nikos Mavrogiannopoulos: gnutls is moving|
As I mentioned before the reason I got interested with licenses again was GitHub. Or more to the point, a presentation by Aaron Williamson  from the Software Freedom Law Center in combination with GitHub's latest changes of adding a license selector.
Aaron Williamson's unscientific study of scraping 28% of the oldest GitHub repositories yielded the disappointing statistic that only ~15% of all repositories had license files, and ~25% of those have the license only mentioned in the Readme file. Out of those licensed repositories the vast majority are either MIT/BSD or Apache 2 licensed. Only about third of all projects where under a Copyleft license.
This general trend with throwing random code into the internet without license declarations is debatable and asks some questions. However I think it shows more that people think licensing is unimportant and only needs a bit of attention more than that people are unaware of the existence of licenses. As such I see GitHub's newly added license choosing helper dialog problematic. When you make a new repository it gives you a dialog to pick a license without any explanation of what the licenses mean. It even bolds some licenses for you. The ones that it deems more important than others are “Apache v2 License”, “GPLv2” and “MIT”. The irony is that two of the licenses that the dialog deems important are actually incompatible with each other. (Apache and GPLv2).
If people did not spend any time before adding a license to their repository they will spend no time thinking about the consequences of licensing. And with all the different forms of the GPL now and all the legal implications that come from it, I am afraid this license selector is going to make things worse rather than better.
When the GPL is involved the complexities of licensing becomes a non fun version of a riddle. So many things to consider and so many interactions to consider.
And that GPL incompatibilities are still an issue that actively effects people is something many appear to forget.
For instance one would think that the incompatibility of the GPLv2 with the Apache Software License 2.0 should be a thing of the past now that everything upgrades to GPLv3, but it turns out that enough people are either stuck with GPLv2 only or do not agree with the GPLv3 that some Apache Software licensed projects are required to migrate. For instance Twitter's Bootstrap is currently migrating from ASL2.0 to MIT precisely because some people still need GPLv2 compatibility. Among those projects that were affected were Drupal, WordPress, Joomla, the MoinMoin Wiki and others. And even that case shows that people don't care that much about licenses any more as Joomla 3 just bundled bootstrap even though they were not licenses in a compatible way (GPLv2 vs ASL 2.0).
The other traditional case of things not being GPL compatible is the OpenSSL project which has a license that does not go well with the GPL. That license is also still incompatible with the GPLv3.
The whole ordeal is particularly interesting as some not so nice parties have started doing license trolling through GPL licenses. The most recent case is Oracle that relicensed Berkeley DB from BSD to APGLv3 which started a lengthy discussion on debian-legal . Primarily because due to that package becoming AGPLv3 it implicitly changes the effective license for 106 other packages to AGPLv3 as well. Considering the license change happens on a Debian installation and not in the original source software this showcases how complicated licenses can become. The original software that depends on the Berkeley DB can remain under its own license just fine if it just always depends on the old version of the library that was BSD licensed. But if Debian would decide to make that software depend on the new version that is AGPLv3 licensed then the whole software would change license to AGPLv3 as well.
|||Rob Landley: GPL version 2 only for BusyBox 1.3.0|
|||Rob Landley: I'm going out now. I may be some time|
|||Rob Landley on Toybox being BSD|
|||Android Licensing Information|
|||LWN: Android and the GPLv2 death penalty|
|||7 Reasons Why Free Software Is Losing Influence|
|||Licensing of Software on GitHub: A Quantitative Analysis by Aaron Williamson|
|||Berkeley DB 6.0 license change to AGPLv3|
You get a headache quite quickly thinking about all these ramifications and reading debian-legal is a weird experience. This gets worse when you think about how people might have different interpretations of the license.
On one hand it's entertaining, on the other it shows you how many people pick licenses with ulterior motives that might not have been written into the original license text. One of my favourite threads on that issue is the one about Nuitka and GPLv3 which at it's core is a discussion if the GPLv3 can be used to make it impossible for others to commercialize the software.
This is especially bad in the GPLv3 which picks specific usecases that were relevant at the time the license was written. The license might now be absolutely useless for new devices that will come up in a few years. At that point it might be impossible to relicense the software because some contributors might disagree or are no longer available. This has been a problem for many years already. MariaDB had to rewrite libmysql because it was GPL licensed. Xapian is trying for years to get GPL code out of the codebase as the original copyright holders are no longer interested in the project or unavailable.
What if it will become illegal to sell consumer devices without signature checks for all software running on it. What if certification will be required for system software? The GPLv3 is already pretty irrelevant in many areas as software companies have figured out that rewriting is less of an issue than license fulfilment.
It's not just GPL though that is a problem here. The Apache Software License is quite a mouthful as well and I am pretty sure that not everybody that licensed code under it read all of the implications of the license.
As you remove text from the license other parts might be coming in. The MIT license is barely two paragraphs and a warranty statement, but how it interacts with local law is not something everybody is aware of. The implicit assumption for many people is that somehow American law applies but that's not always the case. Open Source is international and not every country is the same. Germany and Austria for instance have very few provisions bound to the Copyright itself and don't even provide mechanisms to transfer it. More is bound to the Usage right which the copyright holder can sublicense. Considering that doesn't actually happen in the license declarations I sometimes wonder if someone could hang my software up on that formality.
I believe what's currently happening is something that's new with my generation and that's probably the biggest reason of decline for the GPL going forwards. My generation sees copyrights has a concept that should be much more restricted and have a smaller lifetime. This is interestingly enough exactly what Richard Stallman does not want. He's painfully aware that Copyleft is based on Copyright and as such can only be enforced if there is a strong Copyright behind it.
People that license software under the BSD or MIT license probably would not mind that much if copyrights would be abolished or greatly restricted. Richard Stallman's world on the other hand would would fall apart. He even made a statement about how the Pirate Party will backfire on Free Software .
The new generation has a new view on sharing and money as it stands. They want to make it easy to share content and software and also make it easy at the same time to enable independent monetization. This generation is the generation that goes on youtube and makes reinterpretations of other people's music, that create narrated play-throughs through interactive entertainment media like computer games, that provide lessons involved other people's content etc.
|||Richard Stallman: How the Swedish Pirate Party Platform Backfires on Free Software|
Will this general approach cause problems? Probably. The current state of licensing and license ignoring on GitHub is probably a good indication that there will be problems. But I think we should start to seriously consider simplifying our software licensing environment as we otherwise will have no ideas what will happen in a few years from now.
I think at that point it would be interesting to think about how to explain people in the clearest possible way way the implications of software licenses are and how they can reach their goal. And that would include some flow graphs that point out the problems with cross license compatibility, what the lack of a contributor statement could mean, what happens if copyright holders die or become unavailable and similar things.
I am pretty sure some clever UX designer could make that into an engaging experience that gives people a licensing 101 in 10 minutes that includes all important details in an interesting way. But it would have to be backed by information that an actual lawyer checked together with members of the community to come to some conclusions about what it means for the ecosystem.
Right now I believe, the license choosing wizard of GitHub is a shitty “solution” to the problem of people not adding licenses to their code. And it's probably not just a bad solution, it's one that might actually be hurtful if people are not aware of what the effects of the respective license are.