Information Management: A Proposal

The situation

In 1989 CERN, the European laboratory for particle physics near Geneva, employed thousands of people, many on short visits, and hosted many more visiting scientists from around the world. They used a bewildering variety of computers and operating systems. Documentation about experiments, software, equipment and people was scattered across these systems in incompatible formats, and much of it lived only in people’s heads.

Tim Berners-Lee, a British software engineer at CERN, had experienced the problem directly. Years earlier, during a short contract at CERN, he had written a personal program called Enquire to keep track of connections between people, programs and projects. In March 1989 he wrote a proposal to his managers.

What it argues

The proposal begins with the problem. CERN is a wonderful organisation, it says, but its high turnover means knowledge leaks away: “When two years is a typical length of stay, information is constantly being lost.” New arrivals spend time finding things out that others already know; when people leave, what they knew often goes with them.

It then argues that existing systems fail because they impose the wrong structure. Many documentation systems are organised as trees, hierarchies of categories. Trees are useful, the proposal concedes, but a hierarchical system “does not allow the system to model the real world,” where connections cross categories in every direction. “This is why a ‘web’ of notes with links (like references) between them is far more useful than a fixed hierarchical system.”

The proposal sets out requirements. Remote access across networks. Heterogeneity, so the system works on CERN’s many kinds of computer. And, under the heading of non-centralisation: “A new system must allow existing systems to be linked together without requiring any central control or coordination.” Anyone should be able to add information and links. The system should be simple enough that people would actually use it.

It is notably modest about some issues. Discussions of hypertext, it observes, have sometimes tackled copyright enforcement and data security, but “these are of secondary importance at CERN, where information exchange is still more important than secrecy.”

The idea underneath

The proposal’s core insight is about the shape of knowledge. Hierarchies are a form of legibility: they make information tidy and easy to administer, but they force a messy, interconnected reality into categories it does not fit. A web of links follows the actual connections, at the cost of tidiness. This is Vannevar Bush’s associative principle from “As We May Think”, applied to an organisation.

Its second insight is about dispersed knowledge. Knowledge at CERN was scattered among thousands of people and many systems. A central database maintained by a central authority could never capture it. A decentralised system in which anyone could link anything could at least let the knowledge connect itself. This is also why the proposal’s design fits the end-to-end philosophy of the internet it ran on: keep the core simple and let the edges do the work.

The decision to give up guarantees, links might break and nobody was in charge, was the deepest strategic choice. It is exactly where the web departed from Xanadu, and why it spread.

How it was received and what followed

The proposal did not receive immediate formal approval; Berners-Lee’s manager, Mike Sendall, reportedly wrote “vague but exciting” on the cover and allowed him to work on it. In 1990, with Robert Cailliau, Berners-Lee revised the proposal and wrote the first web server, browser and editor on a NeXT computer. The first website went online at CERN. In April 1993 CERN released the web software into the public domain, royalty-free, and the Mosaic browser, released the same year, made the web easy to use. Growth was explosive. Berners-Lee founded the World Wide Web Consortium in 1994.

The proposal’s remark that copyright and security were secondary concerns was correct for CERN and the opposite of what happened as the web spread to the world. Security, privacy, trust, spam and paying creators became central and largely unsolved problems, discussed throughout this site.

Reading it today as a decision-maker

The proposal is a model of how a world-changing idea can begin with a local, concrete problem. Berners-Lee did not propose to reorganise the world’s knowledge. He proposed to stop CERN losing track of what it knew. The local problem gave the idea early users and a reason to exist; the general design let it escape.

It also shows the cost of decisions that are right for the first context and wrong for the second. Choosing openness over security made perfect sense in a community of physicists sharing information. It created enormous problems when the same design served billions of strangers. Ask of any design which assumptions belong to its first users, and what will happen if the product outgrows them.

Sources and limitations

Quotations are from the copy of the proposal published by the W3C and were checked against it. The “vague but exciting” annotation and the history of the first server are drawn from Berners-Lee’s own account in Weaving the Web and from CERN’s history pages. The interpretation in terms of legibility and dispersed knowledge is the author’s.