At 14:56:20 GMT on 6 August 1991, Tim Berners-Lee answered a stranger's question. Nari Kannan, posting from a Digital Equipment Corporation address, had asked the alt.hypertext newsgroup four days earlier whether anyone knew of research or development into "hypertext links enabling retrieval from multiple heterogeneous sources of information." Berners-Lee replied that yes, there was a project, and it was called WorldWideWeb, and the address format included an access method so a link could point at almost anything. He signed off with "Collaborators welcome! I'll post a short summary as a separate article." Just over an hour later, he did.

Today gets marked as the launch of the world's first website, and that framing is off by seven and a half months. The site was already there. CERN's own timeline puts the first browser, editor, server and website live by Christmas 1990, and the first pages describing the project went up on 20 December 1990. By January 1991 there were web servers running outside CERN. In March the line-mode browser reached a limited audience on a handful of machines, and on 17 May it went into general release across CERN's central computers.

None of that was a project in the sense of having a budget. Berners-Lee took the proposal to his supervisor Mike Sendall in March 1989 and got, by most accounts, a fairly tepid response. CERN never funded the work as such. What Sendall eventually granted, around September 1990, was time rather than money: permission to build the thing in the gaps of a job about particle physics. That constraint shows up all over the design.

What changed on 6 August wasn't that something came into existence. It was the composition of the audience. Everything up to that point had travelled along institutional lines, to people who worked at CERN or at a lab that collaborated with CERN, and the software followed the relationships. From that Tuesday the software could be fetched by anyone who happened to read a hypertext newsgroup, with no introduction and nobody to ask. The project history page Berners-Lee maintained, which is still sitting at info.cern.ch in its original markup, records the month like this: "Files available on the net, posted on alt.hypertext (6, 16, 19th Aug), comp.sys.next (20th), comp.text.sgml and comp.mail.multi-media (22nd)." Six posts across five newsgroups over sixteen days, one man working methodically through a list, going where the people who might care were already sitting and rewording it slightly for each new crowd.

The part of the summary post that did the work is not the prose. It's a file path. Buried under a heading that just says "Try it," the executive summary gives the node as info.cern.ch, with the raw IP address in square brackets next to it because you couldn't assume name resolution would find it, and then the file: /pub/WWW/WWWLineMode_0.9.tar.Z. Version 0.9. He calls it a "prototype (very alpha test) simple line mode browser." Everything that followed came out of an anonymous FTP directory and a version number that hadn't reached 1 yet.

The writing around it is worth reading for tone alone, because there isn't any. "The WWW project merges the techniques of information retrieval and hypertext to make an easy but powerful global information system." It reads like the abstract of a grant application, which is more or less what it was. The one line that carries any heat is the second: "The project started with the philosophy that much academic information should be freely available to anyone." That's an argument about who gets to read things, filed as a description of scope.

Then there's the design decision sitting near the end of the summary. "Making a web is as simple as writing a few SGML files which point to your existing data," he writes, and then: "The very small start-up effort is designed to allow small contributions." Designed to. Not happened to allow, not turned out to be easy. Someone with no budget had thought about the smallest useful thing a stranger could do and made sure that thing stayed cheap.

The shorthand version of this story usually says Berners-Lee refused to patent the web, and holds that up as the moral centre of the whole thing. The instinct is right and the mechanism was slower and duller than the telling suggests. What actually made the web free was a piece of paperwork: on 30 April 1993, nearly two years after the newsgroup posts, CERN relinquished all intellectual property rights to the code, source and binary alike, and gave anyone permission to use, duplicate, modify and distribute it. I've written before about how almost nobody noticed at the time. The generosity was real, and it still had to pass through an institution's paperwork before it counted for anything.

The NeXT cube running the first server sat in Berners-Lee's office with a label on it, written by hand in red ink, reading DO NOT POWER IT DOWN!!. Two exclamation marks. The entire World Wide Web, at that point, was one machine somebody might unplug to hoover.

Which is roughly the problem Nicola Pellow solved. The browser in that tarball, the line-mode client, was hers, written during a student placement, and it ran on ordinary terminals rather than on the expensive NeXT hardware Berners-Lee had developed against. Without it the thing Kannan was being invited to try would have been untryable for almost everyone who read the post. The cheap-entry principle in the executive summary is only a principle until someone writes the client that makes it true.

Thirty-five years on, both posts from that Tuesday are still served from w3.org with their Usenet headers intact, message IDs and all, which is why the second one can be dated to 16:00:12 GMT rather than merely to August. That precision is unusual for 1991, and mostly an accident of the people involved being the sort who kept logs. I've argued before that nobody can correct the record where nobody kept one.

Kannan asked a technical question and got a technical answer. The replies that followed were about link qualifiers.

Sources: