One sheet of paper with two signatures, issued on the last day of April 1993, relinquished Europe’s largest physics laboratory’s rights to a technology that would reshape the global economy ten years later. But to understand exactly what CERN gave up, we first need to understand what it created — and what it did not.
In March 1989, Tim Berners-Lee, a thirty-three-year-old British programmer working at CERN on contract, placed a document with the dry title “Information Management: A Proposal” on his manager Mike Sendall’s desk. The document addressed a specific and rather mundane problem: in a laboratory where thousands of physicists from dozens of countries worked at once, information was getting lost faster than it was being generated. Telephone directories went out of date within months, experiment documentation existed in incompatible formats on different machines, and every new employee spent weeks figuring out who was responsible for a particular detector and where its specifications were kept. Berners-Lee’s proposal described a system of hypertext links that would connect documents across the existing network — without a central catalog, without a single point of control. Sendall wrote a note in pencil on the cover that became perhaps the most famous marginal comment in the history of technology: “Vague, but exciting.”
This note was neither project approval nor a budget allocation. It meant something like permission to spend work time on a prototype. Sendall, colleagues recalled, had a knack for spotting promise in fuzzy sketches — and for shielding people working on things outside the laboratory’s direct mandate. CERN existed for particle physics, not for developing communications protocols. But CERN had a long-standing tradition: side tools built for experiments sometimes turned out to be more valuable than the experiments themselves. Detector technologies became medical scanners; data-processing methods became statistical packages. Berners-Lee was essentially proposing another such tool, only no one could foresee the scale of its consequences — including Berners-Lee himself.
By the end of 1990, Berners-Lee and the Belgian engineer Robert Cailliau had assembled a working prototype. Its hardware foundation was a NeXT computer — an elegant black cube designed by Steve Jobs’s company after he left Apple. The machine was expensive and rare, but it had a sophisticated development environment that made it possible to build graphical applications quickly. This cube ran both the world’s first web server and the first web browser, named WorldWideWeb (later renamed Nexus to avoid confusion with the system itself). The browser could both display pages and edit them — it was closer to modern wiki systems than to Chrome or Firefox.
Alongside this browser came the line-mode browser — a text client that ran on any terminal connected to a Unix machine. It was this that made the web accessible beyond the circle of lucky NeXT owners. If WorldWideWeb on the NeXT cube was a showroom, demonstrating what the web could be, the line-mode browser was the workhorse, proving that the web already worked. The third component was the libwww library — a set of shared code that other developers could use to create their own clients and servers. Roughly speaking, Berners-Lee released not a finished product but a box of building blocks: the hypertext transfer protocol (HTTP), the markup language (HTML), the addressing system (URL), a working server, two browsers, and a library for anyone who wanted to build their own.
By 1991, the first servers outside CERN had already appeared — mostly in other physics laboratories and universities. But adoption was slow, and one reason was uncertainty: on what terms could all this be used? CERN is an international organization with a special legal status, member states, and obligations to the taxpayers of twenty countries. Any university considering setting up a web server might reasonably wonder: would a letter arrive from Geneva a year later demanding licensing fees?
The document dated April 30, 1993, signed by Walter Hoogland and Helmut Weber, was brief and specific. CERN fully relinquished its rights — to both source code and object code — for the server software, browsers, and the libwww library. The formula was radical: public domain — a complete, unconditional waiver of copyright, without conditions or caveats.
I like the precision of this decision, because that is where its power is hidden. CERN did not give away “the internet” or “the web” in some abstract sense. It gave away specific programs: specific lines of code in specific languages. The HTTP and HTML protocols are specifications, descriptions of formats, and they were not anyone’s property in the same way software code was. URL as an addressing system is not a program either, but a convention. Standards and protocols exist in a different legal space: they cannot be “given away,” because by their nature they are closer to ideas than products. Patents are a separate matter, and CERN did not claim patents on core web technologies, though in theory it could have tried.
The popular line that “Tim Berners-Lee gave the internet away for free” is beautiful, but inaccurate three times over. First, the decision was made by CERN as an organization — Berners-Lee was the technology’s author, but his employer was its legal owner. Second, “the internet” and “the web” are different things: the internet, a network based on TCP/IP, had existed since the early 1980s and never belonged to CERN. Third, what was given away were specific programs — the idea of hypertext navigation and the infrastructure of the global network had nothing to do with this act. But even with all those caveats, it is hard to overstate the importance of the April 30 document — and here is why.
CERN did not build a business around software licenses. The laboratory had no sales department, no marketing strategy, no plans to monetize web technologies. Its budget came from member-state contributions, and its mission was framed around fundamental physics, not the commercialization of side products. In a sense, CERN was the ideal candidate to relinquish its rights: it did not need them.
But that explanation is only half the picture. The other half is more pragmatic. By 1993, it had become clear that the web could grow into something far bigger than an internal tool for physicists — but only if others took it up. Every server installed at a university, every new browser written by an enthusiast, increased the value of the entire network. The network effect — the mechanism by which a system’s usefulness grows faster than its number of participants — works only when the barrier to entry is low. Licensing fees, even symbolic ones, create friction. Legal uncertainty creates even more friction: corporate lawyers would rather abandon a technology than untangle the licensing terms of an international organization with twenty member states.
There was an alternative: CERN could have gone down the licensing route, as some universities did with software. But by then competing hypertext systems had already emerged — Gopher from the University of Minnesota, for example, was gaining popularity faster than the early web. And this is where an episode occurred that is often cited as a warning: in 1993, the University of Minnesota announced the possible introduction of licensing fees for Gopher server software. The community’s reaction was immediate and devastating: administrators began migrating to the web, which CERN had just made completely free. Gopher’s demise as a mass system cannot be explained by that episode alone — the web was technically more flexible — but the licensing threat accelerated the exodus. CERN could see what was happening to Gopher, and it vividly confirmed the wisdom of the course it had chosen.
A full relinquishment of rights is an elegant decision, but a vulnerable one. If code is in the public domain, any company can take it, modify it, and close it off — turn it into a proprietary product with no obligation to return its changes to the community. This was precisely what concerned François Flückiger, who took over coordination of web development inside CERN after Berners-Lee left for the W3C in 1994. On November 15, 1994, Flückiger announced to the community the release of version 3.0 of CERN’s server software under the laboratory’s first open license. The license was permissive — allowing the use, modification, and distribution of the code — but retained CERN’s copyright. This was a subtle but fundamental move: copyright made it possible to pursue anyone who violated the license’s terms, that is, anyone who tried to close off the code.
Remarkably, it took only a year and a half to discover the vulnerability in the strategy of total openness and adjust course. In July 1995, MIT, which had become the W3C’s institutional home, released version 3.1 — based on CERN’s version — under a similar permissive license, cementing the practice. This model — “take it, use it, modify it, but do not pretend you wrote it and do not close off the license” — became a template for many later open licenses. In essence, CERN traveled in two years a path that took the free software movement a decade to think through: from the naïve generosity of the public domain to the conscious protection of openness through legal tools.
CERN’s decision of April 30, 1993, did not create the modern web. It removed one — critically important — barrier from the path of a technology that needed dozens of other conditions to take off. By then, the TCP/IP-based internet had been connecting universities and research centers for more than ten years. The DNS domain name system had been operating since the mid-1980s. Personal computers sat on millions of desks. Without all this, CERN’s web server would simply have had nothing to connect to.
The Mosaic browser, created in the National Center for Supercomputing Applications (NCSA) at the University of Illinois, appeared in that same 1993 and did more to popularize the web than the original WorldWideWeb on the NeXT — because it ran on Windows and Mac, had a graphical interface capable of displaying images directly in the body of a page, and was easy for non-specialists to use. Marc Andreessen and Eric Bina, who wrote Mosaic, used the open HTTP and HTML specifications and could do so without looking over their shoulders for CERN’s licensing restrictions — precisely because the April document had settled every question. Andreessen later founded Netscape, and then came the browser race, commercialization, advertising models, search engines — the entire avalanche we retrospectively call “the internet revolution of the 1990s.”
Standardization was not automatic either. The HTTP and HTML specifications developed through the IETF (Internet Engineering Task Force) and then through the W3C, founded by Berners-Lee at MIT in October 1994. The W3C operated on a consensus model: member companies — from IBM to Microsoft — took part in developing standards and committed not to patent technologies incorporated into the consortium’s recommendations. This created a legal perimeter around the web’s core protocols: even if an individual company could obtain a patent on a particular implementation, the standards themselves remained open. The mechanism was imperfect — the browser wars of the late 1990s showed that open standards did not stop companies from creating incompatible extensions — but it prevented the worst-case scenario, in which the web’s core protocols became someone’s intellectual property.
One paradox honestly puzzles me: a decision made in the name of openness ultimately created the platform for the most closed and monopolistic businesses in history. Google, Facebook, Amazon — all of them are built atop the web’s open protocols. They pay no royalties for HTTP. They do not license HTML. They freely use the infrastructure CERN placed in the public domain — and build walled gardens on it, from which user data does not make its way back out. In recent years, Berners-Lee has repeatedly said that the web did not develop quite as he had hoped, and he has promoted the Solid project — an attempt to return control over their data to users.
But that does not mean openness was a mistake. A closed web — a web owned by CERN or licensed through it — most likely would not have become a web at all. It would have remained a tool for physicists, as the 1989 proposal had intended. Or another system would have displaced it — perhaps proprietary, perhaps belonging to a single company. CompuServe, AOL, Prodigy — in the early 1990s, all of them were building their own closed network services and seemed to be the future of online communications. The open web defeated them not because it was technically superior, but because anyone could set up a server, anyone could write a browser, anyone could create a page. Zero entry cost produced exponential growth, and exponential growth made closed alternatives meaningless.
The April 30, 1993 document — two signatories, several paragraphs of legal text — did not “give the world the internet.” It removed the lock from a door behind which stood one specific technology, created for the needs of one specific laboratory. What the world did with that technology afterward is another, far longer and more tangled story. But the door opened on that day, and no lock was ever put back on it. The original document, incidentally, is still kept in CERN’s archives — and is still available via a link on the laboratory’s website, which is itself a small demonstration of how the technology whose rights it waived actually works.