Written by: Vitalik Buterin
Compiled by: Saoirse, Foresight News
In the field of free open source software (and more broadly, free content), copyright licenses are mainly divided into two categories:
If content is released under a permissive license (such as CC0, MIT), anyone can access, use, and redistribute that content without restriction, as long as they comply with the minimum requirement to attribute the source;
If content is released under a copyleft license (such as CC-BY-SA, GPL), anyone can similarly access, use, and redistribute copies without restriction, but if they create and distribute derivative works by modifying or combining with other works, the new work must be released under the same license. Additionally, the GPL requires any derivative works to publicly disclose their source code and other requirements.
In short: permissive licenses allow free sharing with everyone, while copyleft licenses share only with those who are also willing to share freely.
Since I became aware, I have been an enthusiast and developer of free open source software and free content, passionate about building things I believe are useful to others. In the past, I favored permissive licensing models (for example, my blog adopts the WTFPL license), but recently I have gradually shifted to support the copyleft model. This article will explain the reasons for this shift.
The WTFPL advocates a concept of software freedom, but it is not the only paradigm.
Why I once favored permissive licenses
First, I hope to maximize the adoption rate and dissemination range of my works, and permissive licenses clearly state: anyone creating based on my work does not need to worry about any restrictions, which conveniently facilitates this. Companies are often unwilling to open source projects for free, and I know I cannot push them to fully transition to the free software camp, so I hope to avoid unnecessary conflicts with their established and unwilling-to-give-up models.
Secondly, from a philosophical perspective, I generally dislike copyright (and patents). I do not agree with the notion that two people sharing data fragments privately could be seen as a crime against a third party. They have neither touched nor interacted with the third party and have not deprived it of any rights (it should be noted that 'not paying' is different from 'theft'). Due to various legal factors, explicitly releasing works into the public domain is operationally quite complex. Permissive licenses offer the purest and safest way to come as close as possible to a 'no copyright claim' state.
I truly appreciate the concept of copyleft 'using copyright to enforce copyright', which I find to be a clever legal idea. In a sense, it resonates with my philosophical admiration for liberalism. As a political philosophy, liberalism is often interpreted as: prohibiting the use of any violence except to protect people from harm. As a social philosophical idea, I often see it as a way to tame humanity's aversion to harmful reflexes, venerating freedom itself as sacred, making any act that tarnishes freedom become something repugnant; even if you find the unconventional relationships between individuals uncomfortable, you cannot pursue them, as interfering in the private lives of free individuals is itself detestable. Therefore, in principle, history has shown that the aversion to copyright and the practice of 'using copyright to enforce copyright' can coexist.
However, although copyleft for written works fits this definition, GPL-style code copyright has transcended the minimalist concept of 'using copyright to enforce copyright': it uses copyright for the aggressive purpose of 'forcing the disclosure of source code'. While this move is made in the public interest rather than for personal profit, it still constitutes an aggressive use of copyright. This situation is even more pronounced for stricter licenses like AGPL: even if derivative works are only provided via software as a service (SaaS) and never made public, they are still required to disclose their source code.
Different types of software licenses set different conditions for sharing the source code of derivative works. Some licenses require the source code to be made public in a wide range of scenarios.
Why I now favor copyleft
I shifted from favoring permissive licenses to supporting copyleft due to two major industry transformations and a philosophical change.
Firstly, open source has become mainstream, making it more feasible to encourage companies to embrace open source. Nowadays, numerous companies across various industries are embracing open source: tech giants like Google, Microsoft, and Huawei not only accept open source but also lead the development of open source software; emerging fields like artificial intelligence and cryptocurrency rely on open source more than any previous industry.
Secondly, competition in the encryption field is becoming increasingly fierce and profit-driven, and we can no longer simply expect people to open source out of goodwill. Therefore, promoting open source cannot rely solely on moral appeals (such as 'please open the code'), but also needs the 'hard constraints' of copyleft, opening code permissions only to developers who are also open source.
To visually present how these two forces enhance the relative value of copyleft in a chart, it can be roughly illustrated as follows:
In a situation that is neither completely impractical nor necessarily feasible, the value of incentivizing open source is most significant. Today, the mainstream corporate field and the encryption industry are in this state, greatly enhancing the value of incentivizing open source through copyleft.
(Note: The horizontal axis represents the level of motivation for transitioning to open source, while the vertical axis represents the probability of open sourcing. A comparison of the two charts shows that the current mainstream fields find it easier to synergize motivations and effects in promoting open source through copyleft, while the encryption field faces diminishing marginal returns due to ecological maturity, reflecting that the value logic of copyleft incentivizing open source changes with industry development.)
Thirdly, Glen Weyl's economic theory convinces me: in the presence of super-linear returns to scale, optimal policy is not the strictly property rights system of Rothbard/Mises. On the contrary, optimal policy indeed requires a certain degree of proactive promotion of projects to make them more open than they would be otherwise.
Fundamentally, if we assume the existence of economies of scale, it can be understood through simple mathematical reasoning that a non-zero degree of openness is the only way to prevent the world from ultimately becoming controlled by a single entity. Economies of scale mean that if the resources I have are twice as much as yours, my progress will exceed double. As a result, by next year, my resources may become 2.02 times yours, and so on...
Left chart: Proportional growth model, slight differences in the early stages still maintain a small gap; Right chart: Economies of scale growth model, slight differences in the early stages evolve into huge gaps over time.
From a historical perspective, a key factor in curbing the loss of balance is that humanity cannot escape the diffusion effect of progress. Talent carries ideas and skills when flowing between enterprises and countries; poor countries can achieve catch-up growth through trade with rich countries; industrial espionage is common, making innovation difficult to be absolutely monopolized.
However, in recent years, multiple trends are threatening this balance while weakening the traditional factors that curb imbalanced growth:
Technological progress is accelerating at a super-exponential rate, with the speed of innovation and iteration far exceeding the past;
The political instability within and between countries is worsening: if the rights protection mechanisms are well-established, the rise of others does not directly pose a threat; but in an environment where coercive actions are more likely to occur and difficult to predict, excessive dominance by a certain entity can become a substantive risk. Meanwhile, governments worldwide are less willing to regulate monopolies than before;
Modern software and hardware products possess the capability of closure: traditional product delivery necessarily accompanies technological transparency (such as reverse engineering), but today closed-source products can only open usage rights while retaining modification and control rights;
The natural limits of economies of scale have been weakened: historically, large organizations were limited by high management costs and difficulties in meeting localization demands, whereas digital technology has made super-large-scale control systems possible.
The changes mentioned above have intensified the ongoing and even self-reinforcing power imbalance between enterprises and nations.
Therefore, I increasingly agree on the need to proactively incentivize or enforce the diffusion of technology through stronger measures.
Recent policies from various governments can be seen as coercive interventions in technological diffusion:
EU standardization directives (such as the latest mandatory USB-C interface) aim to dismantle closed ecosystems that are incompatible with other technologies;
China's mandatory technology transfer rules;
The U.S. ban on non-compete agreements (I support this policy as it forces companies to achieve 'partial openness' through talent mobility, despite the existence of confidentiality agreements, the actual enforcement is full of loopholes).
In my view, the drawbacks of such policies often stem from the nature of their government-imposed policies, which leads them to prioritize incentivizing those types of dissemination that heavily lean toward local political and business interests. However, the advantage of such policies is that they can indeed stimulate higher levels of technological dissemination.
Copyleft constructs a vast pool of resources (code or other creative works) that can only be used legally when users are willing to share the source code of the content developed based on these resources. Therefore, copyleft can be seen as a highly universal and neutral mechanism for incentivizing technological diffusion, capable of reaping the positive effects of the aforementioned policies while avoiding many of their drawbacks. This is because copyleft does not favor any specific entity and does not require central planners to actively set parameters.
These views are not absolute. In scenarios of 'maximizing dissemination', permissive licenses still have value. But overall, the comprehensive benefits of copyleft have far exceeded those from 15 years ago. Projects that chose permissive licenses back then should at least consider transitioning to copyleft now.
Unfortunately, the meaning represented by the 'open source' label today is completely unrelated to its original intent. However, in the future, we might have open source cars, and copyleft hardware could potentially help realize this vision.
