The Open Source Movement

Open-source software runs the internet, the world’s stock exchanges, every Android phone, most of the cloud, and a growing share of the artificial-intelligence stack. The story of how a movement of volunteer programmers, starting in the 1980s, came to dominate an industry is one of the most important and least understood in the history of technology. It is also a story in which the meaning of the words has shifted several times, and in which the people most often credited with the movement — Richard Stallman, Linus Torvalds — disagree with each other about what it actually means.

The 1970s: software as a shared artefact

The history of open source begins, in a real sense, before software was a product. In the 1950s and 1960s, software was written by researchers at universities, at government labs, and at a few large companies (IBM, Bell Labs, GE). It was distributed with the hardware, or shared as printed listings, and was routinely modified, ported, and exchanged between labs. The commercial value of software was negligible; the commercial value of the hardware that ran it was enormous. IBM famously unbundled its software from its hardware only in 1969, and only under pressure from the US Department of Justice antitrust case. AT&T was prohibited by the 1956 consent decree from selling software commercially, and so distributed the Unix operating system to universities for the cost of the tape and the manual.[1, 2, 16]

The Unix distribution policy had two lasting effects. First, it created a generation of computer scientists trained on Unix, and accustomed to having the source code available, who went on to populate the research labs of the 1970s and 1980s. Second, it created a culture of sharing that the early Unix developers, including Ken Thompson, Dennis Ritchie, and Brian Kernighan at Bell Labs, took for granted. The Unix tradition of small tools, of textual interfaces, of pipes, of shell scripting, and of source-code distribution was a direct product of that culture.[2, 3, 17]

The transition to proprietary software happened gradually in the late 1970s and early 1980s. The 1976 US Copyright Act explicitly extended copyright to software. Microsoft, founded in 1975, was one of the first companies to sell software as a product. AT&T was released from the 1956 consent decree in 1984, and promptly closed the Unix source code, beginning the long series of Unix wars that would eventually give rise to the open-source Unix-like systems. By the mid-1980s, the academic culture of software sharing was under serious threat, and the foundations were being laid for the free-software and open-source movements.[1, 3, 17]

Stallman and the GNU project

The figure most closely associated with the free-software movement is Richard Stallman, a programmer at the MIT Artificial Intelligence Laboratory. In 1983, frustrated by the gradual enclosure of the software commons, Stallman announced the GNU project (a recursive acronym for “GNU’s Not Unix”) with the goal of writing a complete Unix-compatible operating system, distributed under a license that would preserve the freedom to use, study, modify, and redistribute. In 1985, he founded the Free Software Foundation (FSF), and in 1989 he released the first version of the GNU General Public License, the legal instrument that would make the GNU project and its successors possible.[1, 4, 18]

The GNU project was a deliberate act of resistance to the trend toward proprietary software. By 1991, the project had produced a complete Unix-compatible userland: the GNU C compiler (gcc), the GNU Emacs text editor, the GNU core utilities, the GNU Bash shell, the GNU debugger, the GNU C library, and dozens of other programs. The only major piece missing was the kernel. The arrival of Linux in 1991, written by Linus Torvalds, a 21-year-old computer-science student at the University of Helsinki, filled that gap. The combination of GNU userland and Linux kernel became the system now known simply as Linux, although Stallman and the FSF insist on calling it “GNU/Linux”.[5, 6, 19]

Stallman’s contribution was not just code; it was ideology. The GPL, of which the latest version is GPLv3 (2007), is a copyleft license: it uses copyright law to require that any modified version of the code also be released under the GPL, ensuring that the freedoms granted by the license propagate to all downstream users. The contrast with the more permissive BSD and MIT licenses, which allow proprietary derivatives, was, and remains, the central ideological divide in the open-source world. Stallman argued that the freedom of the user was the only legitimate goal; the BSD/MIT camp argued that practical adoption, including by commercial entities, was more important than ideological purity.[4, 8, 20]

Linux and the corporate embrace

Linux, in its early years, was a curiosity. The first versions (0.01 in 1991, 1.0 in 1994) were hobbyist projects, used by a few thousand enthusiasts and a small number of universities. The decisive shift came in the late 1990s and early 2000s, when major commercial users — IBM, Oracle, Red Hat, Intel, and eventually Google and Amazon — began to invest seriously in Linux. The reasons were partly economic (Linux was free), partly strategic (Linux ran on cheap commodity hardware), and partly ideological (the open-source development model produced better software than the closed proprietary model in many domains).[5, 6, 21]

By 2000, Linux was the dominant operating system on web servers, the leading operating system on the world’s top 500 supercomputers (a position it has held continuously since), and the basis of the Android smartphone operating system, launched by Google in 2008. By 2024, Android had more than three billion active users worldwide, making it the most widely used operating system in history. Linux also runs the entirety of the public cloud infrastructure of Amazon Web Services, Google Cloud, and Microsoft Azure, plus a large fraction of the embedded systems in cars, routers, industrial control systems, and consumer electronics.[5, 6, 22]

The development model of Linux is a sustained experiment in distributed, large-scale, voluntary collaboration. Linus Torvalds acts as the benevolent dictator for life, with final say on what goes into the mainline kernel. The bulk of the development is done by employees of commercial companies (Red Hat, Intel, IBM, Google, Samsung, and others) who contribute code back to the upstream project. The contributions are reviewed by a hierarchy of maintainers, with each subsystem (networking, filesystems, drivers, scheduler) having its own maintainer. The model has scaled, remarkably, to tens of thousands of contributors, hundreds of companies, and millions of lines of code, and the resulting kernel is one of the most complex and most reliable software systems ever built.[7, 23, 24]

The Apache Foundation and the open-source web

The other major open-source story of the 1990s was the Apache Software Foundation and the open-source web. The Apache HTTP server, originally written at the National Center for Supercomputing Applications (NCSA) at the University of Illinois, was forked and extended by a group of volunteer webmasters in 1995, and quickly became the most widely used web server on the internet. The Apache License, written in 1995, was a more permissive license than the GPL, and became the dominant license for server-side open-source software.[9, 25]

The Apache Foundation was incorporated in 1999 to provide a legal home for the project and to spawn a host of successor projects: Hadoop for distributed data processing, Kafka for messaging, Spark for in-memory data processing, Cassandra for distributed databases, and many more. The Hadoop ecosystem, in particular, was the foundation of the first generation of modern big-data systems, and many of the most important data engineers of the 2010s learned their craft on it.[10, 25, 26]

The model of the Apache Foundation — corporate-sponsored, vendor-neutral, meritocratic governance, permissive license — has been emulated by a number of other open-source foundations: the Linux Foundation (which now hosts Kubernetes, the Node.js foundation, and many others), the Cloud Native Computing Foundation (which hosts Kubernetes, Prometheus, and others), the Python Software Foundation, the Mozilla Foundation, and many more. The foundation model has become the dominant governance pattern for large open-source projects, particularly in the cloud-native ecosystem.[11, 25]

Open source in the cloud

The cloud computing era has been, perhaps surprisingly, an open-source era. The dominant cloud-native stack — Linux, Kubernetes, Docker/containerd, PostgreSQL, Redis, TensorFlow/PyTorch — is overwhelmingly open source, and the cloud providers have been among the largest contributors. Google open-sourced Kubernetes (originally an internal project called Borg) in 2014; Amazon open-sourced the AWS Nitro System’s hypervisor in 2020; Microsoft acquired GitHub in 2018. The economic logic is straightforward: open source reduces the cost of building the platform, allows the community to debug and harden the software, and creates a workforce trained on the cloud provider’s preferred tools.[10, 11, 27]

The same logic, however, has led to a new set of concerns. The major cloud providers — Amazon, Google, Microsoft — have all been accused of taking open-source projects, building proprietary services on top of them, and contributing back to the community only enough to maintain goodwill. The controversy over Elasticsearch (Amazon’s fork, OpenSearch, after Elastic changed the license to restrict Amazon’s use) is a case in point. The phenomenon — known pejoratively as open-source washing or, more generously, as open-source-led growth — has prompted a wave of new open-source licenses designed to restrict commercial use by hyperscalers. The most prominent is the Server Side Public License (SSPL), introduced by MongoDB in 2018, and the Business Source License (BUSL), used by HashiCorp and others.[12, 27, 28]

Despite these tensions, the open-source cloud stack is now the most important software platform in the world. The Linux kernel alone has more than 30 million lines of code, contributed by more than 4,000 companies and 20,000 individual developers, and runs everything from the smallest embedded device to the largest supercomputer. Kubernetes, the de facto standard for container orchestration, is the second-largest open-source project (after Linux) by number of contributors, and is the foundation of the entire cloud-native ecosystem. The economic value of open-source software, by some estimates, exceeds ten trillion dollars. The economic value of the open-source ecosystem — the companies, the developers, the services, the training, the documentation — is much larger.[11, 27, 29]

The unfinished agenda: open source and AI

The most important open-source question of the 2020s is whether the same pattern of distributed, voluntary, open collaboration can apply to the most consequential technology of the decade: artificial intelligence. The answer, so far, is mixed. The training frameworks — PyTorch (Meta), TensorFlow (Google), JAX (Google) — are open source, and the most popular inference engines, including ONNX Runtime and vLLM, are open source. A growing number of model weights are being released, including Meta’s Llama family, the French startup Mistral‘s models, the Allen Institute’s OLMo, and BigScience‘s BLOOM.[13, 14]

But the most capable models, in 2024, are proprietary. OpenAI has never released the weights of GPT-4 or its successors; Anthropic has never released the weights of Claude; Google DeepMind has released some models (Gemma) but not the most capable ones (Gemini Ultra). The argument for openness, articulated most clearly by Meta’s chief AI scientist Yann LeCun, is that open models are safer (because more people can study them), more innovative (because more people can build on them), and more equitable (because more people can use them). The argument against openness, articulated most clearly by OpenAI and Anthropic, is that the most capable models pose risks of misuse that require controlled deployment. The debate is unresolved, and the outcome will shape the next decade of AI.[14, 15, 30]

The other open frontier is open hardware. The RISC-V instruction set architecture, developed at the University of California, Berkeley, and now stewarded by the non-profit RISC-V International, is the first serious open-source hardware architecture. RISC-V is now backed by a broad coalition of companies (including Google, Intel, Nvidia, Qualcomm, Samsung, and many Chinese chip companies), and is being used in everything from tiny embedded microcontrollers to experimental high-performance processors. The bet — still being tested — is that the same collaborative model that produced Linux can produce competitive processor designs. The trajectory is, in 2024, broadly encouraging.[15, 30, 31]

The open-source movement is, in retrospect, one of the most important and least recognised social phenomena of the late twentieth and early twenty-first centuries. It has produced software that runs almost every important digital system on the planet, has trained multiple generations of programmers, and has provided a working alternative to the closed, proprietary model of software development. Its impact on the cost, the reliability, and the reach of computing has been enormous. Its future, in an age of proprietary AI, is the most important open question in the field.[15, 30, 31]

Comparison: leading open-source foundations (2024)

Foundation	Founded	Headquarters	Notable projects	License model	Estimated budget (2024)
Linux Foundation	2000	San Francisco, USA	Linux kernel, Kubernetes, Node.js, Hyperledger	Mixed (GPL, Apache, BSD)	~$250 million
Apache Software Foundation	1999	Wilmington, Delaware, USA	Apache HTTP Server, Hadoop, Kafka, Spark, Cassandra	Apache License 2.0	~$3 million
Free Software Foundation	1985	Boston, USA	GNU toolchain, GCC, GNU Emacs	GPL (copyleft)	~$1.5 million
Cloud Native Computing Foundation	2015	San Francisco, USA	Kubernetes, Prometheus, Envoy, etcd	Mostly Apache 2.0	Part of Linux Foundation
Python Software Foundation	2001	Beaverton, Oregon, USA	Python, pip, PyPI	PSF License	~$5 million

Key numbers

• ~30 million: lines of code in the Linux kernel (2024).
• ~20,000: individual contributors to the Linux kernel (2024).
• ~4,000: companies that have contributed to the Linux kernel (2024).
• ~3 billion: active Android users worldwide (2024).
• 100%: percentage of the world’s top 500 supercomputers running Linux (2024).
• ~$10 trillion: estimated economic value of open-source software (various estimates, 2024).
• ~90%: estimated share of public cloud workloads running on open-source software (2024).

References

1. Wikipedia, “History of free and open-source software.” (Standard reference on the history of free and open-source software.) https://en.wikipedia.org/wiki/History_of_free_and_open-source_software
2. Wikipedia, “Unix.” (Standard reference on Unix, with a section on the early academic distribution model.) https://en.wikipedia.org/wiki/Unix
3. Wikipedia, “Unix wars.” (Standard reference on the Unix wars of the 1980s and 1990s.) https://en.wikipedia.org/wiki/Unix_wars
4. Wikipedia, “Richard Stallman.” (Standard reference on Richard Stallman and the Free Software Foundation.) https://en.wikipedia.org/wiki/Richard_Stallman
5. Wikipedia, “Linux.” (Standard reference on the Linux kernel and its history.) https://en.wikipedia.org/wiki/Linux
6. Wikipedia, “Linus Torvalds.” (Standard reference on Linus Torvalds.) https://en.wikipedia.org/wiki/Linus_Torvalds
7. Raymond, E. S. The Cathedral and the Bazaar. Sebastopol, CA: O’Reilly Media, 1999. (The classic essay on open-source development models.)
8. Wikipedia, “Open Source Initiative.” (Standard reference on the Open Source Initiative and the history of the term “open source”.) https://en.wikipedia.org/wiki/Open_Source_Initiative
9. Perens, B. “The open source definition.” In Open Sources: Voices from the Open Source Revolution. Sebastopol, CA: O’Reilly Media, 1999. (The original definition of open source, by one of its co-authors.)
10. Wikipedia, “Kubernetes.” (Standard reference on Kubernetes, the dominant container-orchestration system.) https://en.wikipedia.org/wiki/Kubernetes
11. Wikipedia, “Cloud Native Computing Foundation.” (Standard reference on the CNCF and the cloud-native ecosystem.) https://en.wikipedia.org/wiki/Cloud_Native_Computing_Foundation
12. Wikipedia, “Docker (software).” (Standard reference on Docker and the container revolution.) https://en.wikipedia.org/wiki/Docker_(software)
13. Wikipedia, “PyTorch.” (Standard reference on PyTorch, the dominant open-source deep-learning framework.) https://en.wikipedia.org/wiki/PyTorch
14. Wikipedia, “Llama (language model).” (Standard reference on Meta’s Llama family of open-weight language models.) https://en.wikipedia.org/wiki/Llama_(language_model)
15. Wikipedia, “RISC-V.” (Standard reference on the RISC-V open instruction set architecture.) https://en.wikipedia.org/wiki/RISC-V
16. Abbate, J. Recoding Gender: Women’s Changing Participation in Computing. Cambridge, MA: MIT Press, 2012. (A history of computing with attention to the role of women and to the early academic culture of software sharing.)
17. Salus, P. H. A Quarter Century of UNIX. Reading, MA: Addison-Wesley, 1994. (A history of Unix from the inside.)
18. Wikipedia, “Free Software Foundation.” (Standard reference on the FSF.) https://en.wikipedia.org/wiki/Free_Software_Foundation
19. Stallman, R. Free Software, Free Society. Boston, MA: GNU Press, 2002. (A collection of Stallman’s essays on free software.)
20. Wikipedia, “GNU General Public License.” (Standard reference on the GPL, with version history.) https://en.wikipedia.org/wiki/GNU_General_Public_License
21. Wikipedia, “IBM and Linux.” (Standard reference on IBM’s $1 billion investment in Linux in 2000.) https://en.wikipedia.org/wiki/IBM_and_Linux
22. Wikipedia, “Android (operating system).” (Standard reference on Android, the most widely used Linux distribution.) https://en.wikipedia.org/wiki/Android_(operating_system)
23. Wikipedia, “Benevolent dictator for life.” (Standard reference on the BDFL governance model used by Linux and other projects.) https://en.wikipedia.org/wiki/Benevolent_dictator_for_life
24. Corbet, J., and G. Kroah-Hartman. “Linux kernel development.” Annual report, Linux Foundation, 2024. https://www.linuxfoundation.org/
25. Wikipedia, “Apache Software Foundation.” (Standard reference on the Apache Software Foundation.) https://en.wikipedia.org/wiki/Apache_Software_Foundation
26. Wikipedia, “Apache Hadoop.” (Standard reference on Apache Hadoop, the first widely adopted big-data framework.) https://en.wikipedia.org/wiki/Apache_Hadoop
27. Wikipedia, “OpenSearch.” (Standard reference on OpenSearch, the Amazon-led fork of Elasticsearch.) https://en.wikipedia.org/wiki/OpenSearch
28. Wikipedia, “Server Side Public License.” (Standard reference on the SSPL, introduced by MongoDB.) https://en.wikipedia.org/wiki/Server_Side_Public_License
29. Hoffman, C., et al. “The value of open source.” GitHub Octoverse, 2024. (Annual report on open-source software usage and economics.)
30. Wikipedia, “OpenAI.” (Standard reference on OpenAI and its transition from open to closed AI.) https://en.wikipedia.org/wiki/OpenAI
31. Wikipedia, “Mistral AI.” (Standard reference on Mistral AI, the leading European open-weight model developer.) https://en.wikipedia.org/wiki/Mistral_AI