Why has GPL license usage dropped dramatically?

The decline of GPL?

The decline of GPL?
Image by : 

opensource.com

A little while ago I saw an interesting tweet from Stephen O'Grady at RedMonk on the state of open source licensing, including this graph.

Redmonk Black Duck Licensing

This graph shows how license usage has changed from 2010 to 2017. In reading it, it is clear that usage of the GPL 2.0 license, one of the purest copyleft licenses around, has more than halved in usage. According to the chart it would appear that the popularity of open source licensing has subsequently shifted to the MIT and Apache licenses. There has also been a small increase in GPL 3.0 usage.

So, what does all this mean?

Why has GPL 2.0 usage dropped so dramatically with only a marginal increase in GPL 3.0 usage? Why has MIT and Apache usage grown so dramatically?

Of course, there are many interpretations, but my guess is that this is due to the increased growth in open source in business, and a nervousness around the GPL in the commercial world. Let's dig in.

The GPL and business

Now, before I get started, I know I am going to raise the ire of some GPL fans. Before you start yelling at me, I want to be very clear: I am a huge fan and supporter of the GPL.

I have licensed every piece of software I have ever written under the GPL, I have been an active financial supporter of the Free Software Foundation and Software Freedom Conservancy and the work they do, and I advocate for the usage of the GPL. My comments here are not about the validity or the great value of the GPL—it is an unquestionably great license—but more about the perception and interpretation of the license in the industry.

About four years ago, I was at an annual event called the Open Source Think Tank. This event was a small, intimate, annual gathering of executives in the open source industry in the California wine country. The event focused on networking, building alliances, and identifying and addressing industry problems.

At this event, there was a group case study in which the attendees were broken into smaller groups and asked to recommend an open source license for a real-world project that was building a core open source technology. Each group read back their recommendations, and I was surprised to see that every one of the 10 or so groups suggested a permissive license, and not one suggested the GPL.

What matters more is that technology is increasingly becoming open, accessible, and available to everyone.

I had seen an observational trend in the industry towards the Apache and MIT licenses, but this raised a red flag at the time about the understanding, acceptance, and comfort of the GPL in the open source industry.

It seems that in recent years that trend has continued. Aside from the Black Duck research, a license study in GitHub in 2015 found that the MIT license was a dominant choice. Even observationally in my work at XPRIZE (where we chose a license for the Global Learning XPRIZE), and my work as a community leadership consultant, I have seen a similar trend with many of my clients who feel uncomfortable licensing their code under GPL.

With an estimated 65% of companies contributing to open source, there has clearly been a growth in commercial interest and investment since 2010. I believe this, tied with the trends I just outlined, would suggest that the industry does not feel the GPL is generally the right choice for an open source business.

Interfacing community and company

To be honest, GPL's declining popularity is not entirely surprising, and for a few reasons.

Firstly, as the open source industry has evolved, it has become clear that finding the right balance of community engagement and a business model that... y'know... actually works, is a key decision. There was a misconception in the early days of open source that, "If you build it, they will come." Sure, they often came to use your software, but in many cases, "If you built it, they wouldn't necessarily give you any money."

As the years have progressed we have seen various companies, such as Red Hat, Automattic, Docker, Canonical, Digital Ocean, and others, explore different methods of making money in open source. This has included distribution models, services models, open core models, and more. What has become clear is that the traditional software scarcity model doesn't work with open source code; therefore, you need to choose a license that supports the needs of the model the company chooses. Getting this balance between revenue and providing your technology for free is a tough prospect for many.

This is where we see the rub. While the GPL is an open source license, it is fundamentally a free software license. As a free software license, much of the stewardship and support for the GPL has been driven by the Free Software Foundation.

As much as I love the work of the Free Software Foundation, their focus has ultimately been anchored from the perspective that software absolutely has to be 100% free. There isn't much room for compromise with the FSF, and even well-recognized open source projects (such as many Linux distributions) have been deemed "non-free" due to a tiny bit of binary firmware.

This proves complicated for businesses where there is rarely a black and white set of choices and there is instead a multitude of grey. Few businesses share the pure ideology of the Free Software Foundation (or similar groups such as the Software Freedom Conservancy), and thus I suspect businesses are less comfortable about choosing a license that is so connected to such a pure ideology.

Now, to be clear, I don't blame the FSF (and similar organizations such as the SFC) for this. They have a specific mandate and mission focused on building a comprehensive free software commons, and it is perfectly reasonable for them to draw their line in the sand wherever they choose. The FSF and SFC do phenomenal work and I will long continue to be a supporter of them and the many wonderful people who work there. I just believe that a consequence of such purity is that companies may feel uneasy being able to meet the mark, and thus chose to use a different choice of license than the GPL.

I suspect what has also affected GPL usage is a change in dynamic as open source has grown. In the early days, one of the core fundamental reasons why projects would start was a rigorous focus on openness and the ethical elements of software freedom. The GPL was unsurprisingly a natural choice for this projects, with Debian, Ubuntu, Fedora, Linux, and many others as examples.

In recent years though we have seen a newer generation of developers form for whom there is a less critical, and if I dare say it, less religious focus on freedom. For them, open source is a pragmatic and practical component in building software as opposed to an ethical choice, and I suspect this is why we have seen such a growth in the use of MIT and Apache licenses.

The future?

What does this mean for the GPL?

My guess is that the GPL will continue to be a popular choice of license, but developers will view it increasingly as a purer free software license. I suspect that projects that have an ethical commitment to software freedom will prioritize the GPL over other licenses, but for businesses where there needs to be the balance we discussed earlier, I suspect the MIT and Apache licenses will continue to grow in popularity.

Either way, the great news is that open source and free software is growing, and while there may be complexity and change in how licenses are used, what matters more is that technology is increasingly becoming open, accessible, and available to everyone.

13 Comments

Adam Outler

It's only one month into 2017. With only 1/12, even 1/6 of the data in, is it worth analysis?

Vote up!
1
Vote down!
0
Lewis Goddard

It's a comparison of two monthly snapshots, one from 2010, one from 2017. They are therefore comparable data-sets.

Vote up!
2
Vote down!
0
Bradley M. Kuhn

Jono, I think your article above doesn't adequately cover this point: companies and trade associations, in a concerted way, are pressuring developers to avoid copyleft licenses with multiple copyright holders, precisely because they want the opportunity to return to the proprietary licensing scarcity models you mention.

I also urge you to be careful about joining the chorus of people building a Zeitgeist of anti-copyleft sentiment. I know that wasn't your intention, but a lot of people just read headlines: perception becomes reality because truthiness now reigns in our political discourse.

I keynoted at FOSDEM just one week ago about a lot of these issues, so I hope you'll check that out and consider some of the arguments I made there: https://sfconservancy.org/news/2017/feb/13/bkuhn-fosdem-keynote/

The most important point, I think, is that it's not about the orgs you mention (one of which I do work for), but that developers increasingly don't make the decisions about what license to chose: their employers do.

Finally, I note you didn't mention John Sullivan's talk where he does a pretty good job discrediting the so-called "data" that shows GPL is declining -- including discrediting some of the so-called 'studies' you quoted. Here's audio of it: http://faif.us/cast/2012/feb/28/0x23/ and his slides: http://faif.us/cast-media/FaiF_0x23_Is-Copyleft-Framed_slides.pdf

It's important to keep in mind that there is no known statistically valid way to measure license usage in FLOSS, so there isn't actually a way to know factually whether GPL usage is declining. What we can observe is an anecdotal attack by trade associations and companies against copyleft, which includes substantial rhetoric about its decline.

(An interesting example of this: Linux Foundation tweeted just two days ago that a particular package had relicensed from AGPL to Apache license so that "everyone could now use it". LF deleted the tweet after pushback on Twitter, which included the question "When will Linux relicense from the GPL to the Apache license so that 'everyone can finally start using Linux'?". When the organization that holds the name of one of the most popular GPL'd programs has publicly become anti-copyleft like this, we know we have a political attack going on.)

Vote up!
5
Vote down!
0
lightweight

Yup. We have a new generation of devs now starting companies, and they've *always* had Free Software as a raw material. They didn't live through the rough transition from a fully proprietary world where Free and Open Source was treated with open contempt and actively put down by proprietary software vendors. They continue to think that businesses can be "friends" to the community. Some private companies, perhaps, but public corporations, no. This new generation of open source devs put their trust into public corporations like Apple, Microsoft, and Google, because they can't imagine how they could go rogue and close what was once open... This is similar to the conventional wisdom, held by many until very recently, that the US gov't was a benign champion of the citizens of the US and the values of the US' founders... FOSS is insurance for the user, to insure that they, too, can be the developer. It's forgotten by many up-and-coming devs who are "permissive" licensing fans. They forget that "permissive" depends on your perspective. Permissive for businesses, but not for users. I respectfully suggest they reconsider their underlying assumptions.

Vote up!
0
Vote down!
0
bcotton

I agree with you on the observation that there's a new generation of developers who have never known anything but FLOSS as an "acceptable" option. There's some irony, though, in your statement that public corporations can't be friends to the community that is posted on a community-oriented site sponsored by a public corporation.

Vote up!
0
Vote down!
-1
lightweight

No irony... I lay out my rationale for tarring public listed corporations here: https://davelane.nz/megacorps The incentives are crystal clear, and the race to bottom (ethically speaking) is fairly inevitable (there might be some who avoid it, but that'll be the exception not the norm, and who knows which ones).

Vote up!
0
Vote down!
0
jonobacon

"Jono, I think your article above doesn't adequately cover this point: companies and trade associations, in a concerted way, are pressuring developers to avoid copyleft licenses with multiple copyright holders, precisely because they want the opportunity to return to the proprietary licensing scarcity models you mention."

This seems a rather cynical perspective about companies and trade associations that I haven't seen any independently verifiable data to support.

While I don't deny that some companies want to avoid copyleft licensing, from my experience, there isn't pressure on developers to avoid copyleft, it is that copyleft is either (a) not particularly understood within the context of their business model, or (b) copyleft is incompatible with their business model.

"I also urge you to be careful about joining the chorus of people building a Zeitgeist of anti-copyleft sentiment. I know that wasn't your intention, but a lot of people just read headlines: perception becomes reality because truthiness now reigns in our political discourse."

I believe I made it very clear in my article that I am an active supporter of copyleft and the GPL. Sure, some people may only read the headline, but if they are not going to invest the time in reading and understanding the piece, I am not going to invest my time in entertaining their perspectives.

"The most important point, I think, is that it's not about the orgs you mention (one of which I do work for), but that developers increasingly don't make the decisions about what license to chose: their employers do."

This is not particularly surprising, and this is why I think this is a marketing issue. It is entirely reasonable that the company who invests in the work should choose the license. The problem is that the people making those decisions invariably don't understood the GPL and the requirements and benefits therein, and as I mentioned, I would argue that the FSF (and the SFC) has not anchored efforts to resolve this marketing problem (which as I said, I have no issue with - those orgs can use their resources as they see fit.)

"Finally, I note you didn't mention John Sullivan's talk where he does a pretty good job discrediting the so-called "data" that shows GPL is declining -- including discrediting some of the so-called 'studies' you quoted."

Indeed I didn't reference this as I didn't know about it. As John and I discussed on Twitter, I am not saying the data I cited is telling the full story of what is really happening, but I think it is a reasonably credible source that is worthy of us having a conversation, hence my article. Again, while observational data is in no way scientific, it does reflect what I see, and other data sources (e.g. the GitHub license data) suggests that there is indeed a trend occurring.

Also, as I said to John, I would *love* to see another organization provide some additional research. More importantly, I would love to see data behind the argument that GPL licensing is not declining. I am yet to see that data.

"It's important to keep in mind that there is no known statistically valid way to measure license usage in FLOSS, so there isn't actually a way to know factually whether GPL usage is declining. What we can observe is an anecdotal attack by trade associations and companies against copyleft, which includes substantial rhetoric about its decline."

Of course, we can't get a fully accurate picture of this, and I don't think anyone is suggesting that. I don't think we need a fully accurate picture - what is important is that if we sense a trend, which I do believe is happening, we shouldn't bury our heads in the sand, we need to talk about solutions to the issue rather than bickering over the data.

As for your comment about 'an anecdotal attack by trade associations and companies against copyleft, which includes substantial rhetoric about its decline', I just don't buy it. I don't think there is this big conspiracy against copyleft that you see. Of course, I might be wrong, but I would like to see some real independently verifiable data that companies and trade associations are, as a general trend, attacking and encouraging developers away from copyleft. Otherwise, this is just FUD.

Vote up!
0
Vote down!
0
Greg P

Here is some advice. When you are going to submit a graphic as a centerpiece of your article, you should make an attempt to improve its legibility. Even with viewing the graphic separately I can barely read the descriptors if at all. (There are open source tools that could easily edit this). This may be something best presented as a list anyway. Something else which becomes apparent on expanding the graph to legibility is that the top of the graph isn't quite 0.5%, so all of these bars represent very small percentages, and variation from one bar to the next, even comparing 2010 to 2017 is highly exaggerated.

Vote up!
1
Vote down!
-2
jonobacon

Fair point, Greg. The article should at least allow people to click the image and see a bigger version.

Vote up!
0
Vote down!
0
CFWhitman

This data is interesting for certain purposes, but it is a bit misleading when it comes to trying to compare license usage.

It doesn't take into account that the GPL is intended to consolidate efforts while the other licenses are not. The nature of the GPL is that code tends to be contributed to the main project, and even forks usually either get re-absorbed into the main project or become the new main project. There are generally one or perhaps two forks of any project that continue to be maintained while the others wither while giving back any code that was worthwhile to the maintained fork(s).

Related to this is that the data is presented as percentages of the total number of projects rather than the number of users of projects. Because of the nature of the GPL, number of users for their projects tends to go up to much greater levels. Code that is maintained under one of the other licenses only tends to get really big in a project that is either entirely sponsored by a large company or one that ends up treated like a more copyleft project where most code gets consolidated into one big open branch and there are no really significant forks.

Also, the intended use of the project makes a big difference. Libraries for interoperability promoting a standard are much more likely to be released under a permissive license to promote their use in many places, while whole applications have a better chance of being released under a pure copyleft license so they won't be appropriated into a commercial venture.

Vote up!
1
Vote down!
0
Peter Passchier

I think the growing use of MIT-style licenses is due to a lack of awareness of the threats to our freedom. It is very important for the global corpus of software to be increasingly be and stay freely usable, accessible, and modifyable, and the GPL is the best license to guarantee that.

Isn't the main business model based on the MIT-style licenses based on closing the modifications and then monetizing the product (through sales and services)?? I haven't seen that being very sustainably successful yet. While there are very strong successes based on GPL-style licenses.

I understand the sentiments of individuals choosing an MIT-style license, not wanting any encumbrance, "not caring" how their work is used, but I think it is shortsighted and uncaring, not beneficial to them and the community they are contributing to. As for companies going with MIT-style licenses, this is partly giving in to scare-mongering, and partly hope for monetization options in the future that might never play out as they hope.

Vote up!
1
Vote down!
0
simonphipps

I fear the headline (and to some extent the article) reflect the "attack of the compliance-industrial complex" to which Bradley alludes (I heard the term in a talk at FOSDEM so it's not mine!) As he states, there is no evidence the GPL is objectively declining, just that commercial use of open source is increasing. I think it's regrettable this is spun as a strike against the GPL rather than as a win for the whole open source movement.

An open source project is the “safe space” in which developers with an interest in a piece of software, for whatever reason, are able to collaborate over its evolution without the motivations of others impacting their own usage. The open source license the project uses defines and guarantees the things over which those developers need certainty. All give the right to use the code for any purpose, to share it with anyone, and to make whatever changes they want. Beyond those essential freedoms, different communities need different certainties.

Communities working on code that is normally used directly, alone and in its entirety – application software such as LibreOffice, for example – may well want their license to also guarantee reciprocal grants of the same rights to other developers. Most of the LibreOffice core developers work for companies that offer support, training, customisation and deployment. Because everyone who offers these services has to contribute their work as a license requirement, it’s much less likely that a freeloader will be able to undermine their business. The reciprocal license – in this case the Mozilla Public License v2 – is a key attribute expected by the community. Other projects such as GNOME and the Linux kernel use reciprocal licenses (the GPL in those cases) for similar reasons.

This is not true of every community though. For developers mixing ingredients from multiple origins – frameworks, components, libraries – reciprocal license requirements increase the uncertainty rather than decrease it. Their employer may be concerned about managing the different reciprocal duties of different licenses, such as the Eclipse Public License (EPL) and the GPL. The different expectations of the nature and rigor of proof of reciprocity by different communities may also be a concern. For these developers, it is much simpler to use non-reciprocal licenses for their code, especially if the code in question is not directly monetised.

So perhaps a better way to view the subject is to note that the open source world has grown enormously. The use and support of the GPL has also grown with it, but new strengths have also emerged related to corporate adoption of open source. The choices by various communities of the certainties they prefer to protect need respecting.

Vote up!
0
Vote down!
0
Dev Bob

The most important reason for the decline of the copyleft has to do with its fundamental nature - when 2 developers contribute to a copyleft project, both lose the option to use the code (as a whole) outside of a that copyleft, unless each grants special non-copyleft rights to the other. Linux illustrated it when Linus explained why the kernel can't go GPL3 - even the move to another copyleft license became impossible. The larger the project, and the more contributors - the bigger the problem. The added complexity to license maintenance is combined with too many OSS projects chasing too few interested developers, besides devs don't want to monkey around with administrative tasks, so a permissive license gets that off their backs.

Also, in the current dearth of qualified contributors, the contribution of commercial entities has increased and their share can't be ignored. However they don't want the licensing nightmare of of GPL v.X and if a project wants their help, it better be licensed permissively.

Vote up!
0
Vote down!
0

Comment now