The Long Way to This Page

You tap a link, and this page appears. A fingertip touches glass, the phone notices, and within a second or two it turns that touch into radio waves, sends them across the room, sets off a relay through machines owned by people it will never meet, and brings these words back to the same piece of glass. It feels like nothing. It is one of the most complicated things you will do today.

Who invented the machinery that just did it? The question usually gets one name in reply, and the name is usually wrong. Nobody invented the internet. It was assembled over about thirty years by a relay of people in the United States, Britain, France, Australia and elsewhere, and a remarkable number of the decisive pieces were given away rather than sold. The most famous person in the story, an Englishman called Tim Berners-Lee, born in London in 1955, did not invent the internet at all. He invented the web, one of the things that travels over it, and the most important thing that happened to his invention was that nobody ever had to pay for it.

Follow your tap instead, hop by hop, from the glass to the page and back, and stop at each hop to meet the people who built it. The stops come in the order your tap travels, not the order they were invented, so the dates jump about. On the way you will cross, without noticing, the line between the internet and the web.

A Crowded Room Made of Air

The first hop is the shortest: a few metres from your phone to the Wi-Fi router on a shelf. There is no wire, so the phone does what people in a room do. It speaks, in radio waves, and the router listens. Speaking, for a phone, means making rapid, tiny changes to a radio wave in agreed patterns, each change standing for a few of the ones and zeros of your request: Morse code with a torch, only far faster.

The trouble with speaking is that everyone else in the room can speak too. Your laptop, your partner's phone and the television all want the same router on the same slice of radio, and if two of them speak at once, both are garbled. So each device checks that the air is quiet before it talks, and a device that hears someone else waits a random moment and tries again. A slice of radio works like one microphone on a stand: whoever finds it free steps up, and everyone else waits.

The ancestor of that etiquette comes from Hawaii. In June 1971 a team at the University of Hawaii led by Norman Abramson switched on ALOHAnet, linking the university's campuses on different islands by radio. Its rule was cruder than today's: send whenever you have something to send, and if no acknowledgement comes back, wait a random interval and send again. The random wait is the clever part, because two machines that collided are unlikely to pick the same pause again. A plaque put up by the engineers' professional body says ALOHAnet led directly to Ethernet, the cabled networks in every office, and to personal wireless.

Now put the router in a block of flats, where every neighbour has one. Routers work in two stretches of the radio dial, marked 2.4 GHz and 5 GHz on the box; the numbers say how many billion times a second the wave vibrates. The older 2.4 stretch has room for only three slices that do not spill into one another, and the 5 GHz stretch has more but gets through walls worse, the way a neighbour's bass comes through the wall and the singer's voice does not. So the routers of a whole stairwell share a few slices, and every device waits for every other one it can hear, including the ones next door. Your evening Wi-Fi is often not broken. It is waiting for the neighbours to finish.

Microwave ovens share the 2.4 stretch too, because nobody needs a licence to use it: it was set aside for industrial, scientific and medical machines, and in 1985 the American regulator let data radios in as well. Free airwaves are why Wi-Fi is cheap. They are also why the room is crowded.

Walls make it worse in two ways. First, every wall soaks up some of the signal, concrete more than plasterboard. When the signal weakens, phone and router switch to simpler patterns with more repetition built in, easier to make out through the noise but carrying fewer ones and zeros each second. A wall rarely stops Wi-Fi; it slows the conversation, the way you speak slowly and repeat yourself on a bad phone line.

Second, walls throw the signal back. Indoors your phone hears the router many times over: once directly, and again off every wall, cupboard and person. Think of speaking in a cathedral. Talk fast and each syllable drowns in the echo of the one before; talk slowly and the echoes die away between words. Wi-Fi exploits exactly this. Instead of one fast stream, it splits the message into many slow ones sent side by side, each on its own slightly different pitch, like a choir in which every voice holds a separate note: 48 voices carrying data in the first versions to use the method. Each voice is slow enough to shrug off the echoes. Together they are fast.

The choir is older than Wi-Fi: engineers at Bell Labs in America worked it out from 1966, and a 1971 paper showed it could be done cheaply with a calculation called the fast Fourier transform, a quick way of pulling a jumble of radio apart into its separate notes. The people who patented a way to use the choir for fast wireless networks indoors were radio astronomers. At CSIRO, Australia's national science agency, John O'Sullivan and his colleagues had been trying to detect radio signals from tiny exploding black holes, which CSIRO's own film describes as "smaller than atoms". They never found them, but by the early 1990s they had turned their skills on echoes indoors and filed for an Australian patent in 1992, followed by an American one in 1996.

Here the story of gifts has its first exception: CSIRO did not give its invention away. It patented it and sued a string of technology companies in an American court, arguing that its method was built into the Wi-Fi standards, and the cases ended in settlements worth 205 million Australian dollars in 2009 and 220 million in 2012. In all, CSIRO has collected around 430 million Australian dollars from an idea now in more than 15 billion devices. It is a different choice from most of the other hops.

Two popular stories need correcting. The actress Hedy Lamarr really did patent a radio system with a co-inventor: their patent, granted on 11 August 1942, had a torpedo's control signal hop from slice to slice in step with a perforated roll like a player piano's, so the enemy could not jam it. The first standard that Wi-Fi grew from, approved in 1997, even included a hopping option. But Lamarr's design, the American regulator's own expert on such radios later noted, "was not reduced to practice", patent language for never actually built, and modern Wi-Fi runs on the choir instead.

The second is the name. "Wi-Fi" was introduced in 1999 by the industry group that certifies the products, and one of its founders later wrote that the brand agency Interbrand coined it, that "Wi-Fi doesn't stand for anything", and that the tagline "The Standard for Wireless Fidelity" was added afterwards and dropped.

The router now passes your request out of the flat, down a cable and into the rest of the world.

Fifteen Hundred Envelopes

Your request is small. The answer will not be. A web page is heavier than it looks: the median page opened on a phone in 2024 weighed about 2.3 megabytes and needed 66 separate requests for its text, pictures, fonts and code. Your request and the page that comes back are both chopped into small pieces of at most 1,500 bytes each on an ordinary network (a byte is roughly one letter of text), so a typical page returns as something like fifteen hundred little envelopes. Engineers call them packets. Each one carries the address it is going to, the address it came from, and a number saying where it belongs.

Every machine on the way, a router like the box on your shelf but much bigger, looks only at the destination address and passes the envelope to whichever neighbour its map says is the best next step. Envelopes from one conversation usually follow the same road, because routers try to keep them together, but nothing guarantees it: when a road changes, later envelopes take another, and some may arrive out of order. The computer at the far end puts them back in sequence by their numbers and asks again for any that went missing. Real envelopes differ in one way that matters: the sender keeps a copy of everything until the other side confirms it arrived, so a lost envelope costs a moment, not the letter.

That is why the network survives a broken road. The old telephone network reserved a whole line for each call from start to finish, and if a link failed, the call died. A network of envelopes reserves nothing. If a cable or a router fails, then as soon as the routers around it notice, the envelopes go another way.

The idea was invented twice, on two sides of the Atlantic. Paul Baran, a Polish-born engineer at the RAND Corporation, a research institute in California, worried in the early 1960s that America's telephone and military command networks would not survive a nuclear attack, because their central switching centres would be destroyed. His answer, set out in RAND studies from 1960 and collected in one volume in 1964, was a network with no centre, which cut messages into "message blocks" and sent them around any broken part.

In Britain, a Welshman, Donald Davies, born in Treorchy in the Rhondda Valley in 1924, came at the same idea from the other side. Working at the National Physical Laboratory, he noticed that people using a shared computer sent short, irregular bursts of data, while the telephone network was built of fixed circuits meant for steady conversation, so he proposed a network that treated its traffic as short messages. He called the pieces packets, and the internet's founders later wrote that the word was adopted from his laboratory and that the British and American work had gone on in parallel without either side knowing. The Englishman in this story comes later.

Baran's motive is the source of the most repeated story about the internet: that it was built to survive a nuclear war. The network that grew into the internet, ARPANET, was paid for by the Pentagon's research agency, but for a far more ordinary reason. Bob Taylor, one of the men who ran the agency's computer research programme in the 1960s, had three terminals in his office, each wired to a different computer in a different part of the country, and had to change chairs to reach a different colleague; "This is silly," he later told NPR. Taylor also said that ARPANET's creation "was not motivated by considerations of war", and the internet's founders agree: a major motive was resource sharing, letting researchers use scarce, expensive computers elsewhere, and the nuclear story was never true of ARPANET.

Two Letters and a Crash

The first message ever sent across that network went wrong. At 22:30 on 29 October 1969, in a computer laboratory at the University of California, Los Angeles, a student programmer called Charley Kline tried to log in to a computer at the Stanford Research Institute. The word was meant to be "login". Kline got the L and the O across, and then the system crashed. So the first message on what became the internet was, as the professor who ran the laboratory later wrote, "lo", as in "lo and behold", and the full login worked about an hour later.

Your tap does what Kline was trying to do: it asks a distant computer for something. Within two years the network carried email: in 1971 an engineer at the firm that built the network sent the first one between two computers across it, and picked the @ sign for the addresses. Email is nearly twenty years older than the web, which is the simplest proof that the internet and the web are not the same thing.

A Post Office That Never Opens Letters

Your envelopes do not stay on one network. They leave your provider's and cross several others, owned by different companies in different countries, none of which answers to the others. The name is literal: an inter-net, a network of networks. For it to work, every one of them has to agree on the shape of the envelope and on what to do with it.

In 1973 that agreement did not exist, and ARPANET could not talk to the radio and satellite networks the Pentagon was also funding. Bob Kahn, newly arrived at the research agency, asked Vint Cerf, then at Stanford, to help him design a common language. What they built works like the international post. A letter from Tallinn to Osaka passes through postal services that have never heard of you and answer to different governments, and it arrives because each of them accepts an envelope addressed in the agreed way, reads only the outside and passes it on. The routers in the middle of the internet do the same and keep no record of the conversations passing through them; the checking, as you saw, is left to your phone and the server, the computer at the far end that holds the page. Among the ground rules Kahn set himself was that "there would be no global control at the operations level". That is a technical decision with a political effect. Nobody has to direct the internet's traffic for it to run, so nobody does.

In May 1974 the two men published their design as "A Protocol for Packet Network Intercommunication". It was later split into two parts, with names you may have seen in a computer's settings: TCP, the part at the ends that keeps count of the envelopes, and IP, the addressing the post office reads. On 1 January 1983 the network's operators switched the old language off for every computer except a list of stragglers that took most of the year to clear, and buttons were handed out reading "I survived the TCP/IP transition". If someone insists on names for the internet itself, Kahn and Cerf are the nearest thing it has to inventors: they shared the 2004 Turing Award, often called the Nobel Prize of computing, for exactly this work.

Kahn and Cerf did not invent the post-office philosophy alone, and the man they learned from is almost always left out. Louis Pouzin, a French engineer who had worked in Massachusetts on one of the first computers many people could share at once, built a network in France called CYCLADES that was the first to make the computers at the ends responsible for reliable delivery, and demonstrated it in 1973. It sent every envelope out on its own, with no reserved road and no promise that it would arrive. Cerf has recalled that he and Kahn visited Pouzin's team that year, learned about the idea and credited his work in their 1974 paper. When Britain's Queen Elizabeth Prize for Engineering honoured the internet and the web in 2013, it went to five people, Pouzin among them, and not to one.

The Phone Book With No Single Owner

I have skipped a step, the one that actually came first. Routers deliver envelopes to numbers, not names. Every machine on the internet has a numerical address, a phone number for computers. Your phone knew the name of this site, because the link contained it, but before it could address a single envelope it had to find the number that goes with the name. So it asked.

It asked a nearby server, usually run by your internet provider, which works like directory enquiries. If that server did not already know where tsitsejev.com lives, it asked one of the internet's root servers. The root does not know either, but it knows who keeps the list of every name ending in ".com". That keeper does not know the number, but it knows which server holds this site's own entry, and that one gives the number. The nearby server remembers the answers for a while, so the next person to ask gets one at once. Unlike a real phone book, this one is not a single book.

In 1983 every computer on the network was still listed in one file, kept at the same Stanford institute that had received "lo" and copied by everyone, and its size and the frequency of its updates were, as the proposal to replace it put it in November 1983, "near the limit of manageability". That proposal came from Paul Mockapetris, a researcher at the University of Southern California. Instead of one master list, each organisation would keep its own corner of the book. Even the top page, the short list of endings such as .com and .ee, is kept by a single coordinating body and consists mostly of pointers to whoever keeps each ending's list. Nobody holds the whole book, and the book works anyway.

Vague but Exciting

Now your envelopes reach the server, a computer in a data centre of the kind I look after for a living. For the first time on the whole trip, something opens them and reads what is inside. Your browser, the app on your phone that shows web pages, has asked, in a fixed and polite form of words, for one particular page. The server sends back the page itself. Your browser reads it, finds the pictures, fonts and other pieces it lists, and asks for each of them separately, sometimes from other servers. That is where the 66 requests come from.

This is the moment you leave the internet and enter the web. The internet is the roads and the post: the radio hop, the envelopes, the routers, the phone book. The web is one of the things you can reach by those roads, an enormous set of pages linked to one another, just as email and video calls travel the roads without being the web. The web rests on three inventions, and you used all three: an address for every page (the link you tapped), a set of rules for asking for a page and sending it back, and a language in which pages are written, links included. Technical people call them URL, HTTP and HTML. The first versions of all three were defined by Christmas 1990 by one man at CERN, the particle physics laboratory near Geneva.

Tim Berners-Lee had studied physics at Oxford and found that, at CERN, it was often easier to ask people over coffee than to find information on the computers. In March 1989 he gave his boss a proposal for a system that would link documents across the network. The boss wrote on it "Vague, but exciting" and let him carry on. Robert Cailliau, a Belgian engineer at CERN who had proposed a system of linked documents of his own, joined him. Soon the first web server, the computer answering at info.cern.ch, was running at CERN, and in August 1991 Berners-Lee offered the software, on the internet's public discussion boards, to anyone who wanted it.

The decisive act came less than two years later, and it was a legal one. On 30 April 1993 CERN put the basic web software, a program that sends pages out, a simple program for reading them and a kit of parts other programmers could build on, into the public domain with a short statement: "CERN relinquishes all intellectual property rights to this code". In plain terms, anyone could now run a website or write a browser without asking CERN or paying it anything. The document was signed by two of the laboratory's directors, not by Berners-Lee, but he and Cailliau had pushed for it. Berners-Lee credits Cailliau with eighteen months of pressing CERN's management for a royalty-free release. His own reason was practical. A web was only useful if everyone was on it, he wrote in 2025, and he could not also ask "that they pay for each search or upload they made".

The Rival That Charged a Fee

Free can sound sentimental. A rival from Minnesota shows why it was decisive. The link you tapped belongs to the system that won, and in 1993 that was not obvious. Gopher, built at the University of Minnesota, was a simpler system of menus: you picked an item from a list on the screen, then from a list inside that, like working through the drawers of a library card catalogue. That year Gopher was still far more popular than the web, its traffic growing by 997 per cent. Early in the year the university announced that businesses using its Gopher software would have to pay a licence fee of hundreds or thousands of dollars, and by March the Gopher team was answering angry users. Many felt betrayed, above all those who had contributed code in the open spirit of the time. One of the team later put it bluntly: "That socially killed Gopher."

Weeks later, CERN went the other way. Anyone choosing a system to build on could now pick one whose owner wanted money or one whose owner had given up all rights to it. In 1993 web traffic grew by 341,634 per cent, and in the spring of 1994 it overtook Gopher's for the first time. The fee was not the only cause: by the end of 1993 the web also had Mosaic, the first published browser to show pictures alongside the text. But the difference between free for everyone and free for some turned out to matter a great deal.

Berners-Lee does not think the story ended there. In a Guardian essay in September 2025, weeks after his memoir This Is for Everyone, written with Stephen Witt, came out, he asks whether the web is still free and answers: "No, not all of it." He points to a handful of large platforms harvesting users' private data. The gifts kept the roads and the language open to everyone. They did not stop a few companies from building fenced estates on top of them, and much of what you do on your phone now happens inside those fences.

Glass on the Seabed

The page now comes back to you, as envelopes of its own, and if the server is across a sea, part of their journey is under water. "The cloud" lives in data centres on land, but between continents it travels along the seabed. TeleGeography, a firm that maps the cables, checked the much-repeated claim that they carry over 99 per cent of the data travelling between continents and concluded that it is true, though without satellite data it cannot be calculated precisely. For most of its way across an ocean, a cable is about as wide as a garden hose, with your data travelling as pulses of light through glass fibres roughly the thickness of a human hair.

For a reader in Estonia the seabed is close by. In October 2023 the Hong Kong-flagged container ship Newnew Polar Bear dragged its anchor across two telecom cables and the Balticconnector gas pipeline linking Estonia with Finland and Sweden. China later admitted that the ship had damaged the pipeline but blamed a storm. In November 2024 two more cables were cut, and Swedish investigators said it could not be determined with certainty whether a Chinese ship investigated over it had done so on purpose. On Christmas Day 2024 the Estlink 2 power cable between Estonia and Finland was damaged, along with four telecom cables, three of them linked to Estonia, and Finnish investigators said the tanker Eagle S, missing an anchor, had likely caused it. The case against three of the tanker's officers, thrown out by a Helsinki court in October 2025 for lack of jurisdiction, was revived on appeal in August 2026, a ruling that can still go to Finland's Supreme Court. On the last day of 2025 two more cables between Estonia and Finland were damaged, and Finland detained the cargo ship Fitburg, whose anchor chain police found lowered into the sea; its captain and boatswain went on trial in Helsinki in September 2026 and deny the charges.

Each time, the internet in Estonia carried on. After the Christmas breaks, the communications ministry's statement had a plain headline: the cables between Estonia and Finland were disrupted, but communication services were working. Its explanation was just as plain: Estonia's links abroad are duplicated several times over, and when undersea cables fail, other cables, under the sea or overland through Latvia, carry the traffic. It is the detour around a broken road that Baran designed his network for. There are roughly 200 cable faults a year worldwide, and you rarely hear of them because operators spread their traffic across several cables so that the others carry on. The gift runs on a physical body, and the body can be cut.

Back Under the Glass

The page has arrived. In the second or two since your tap, your request used a way of sharing the air first tried in Hawaii and a way of taming echoes that Australian radio astronomers patented. It travelled in envelopes a Welshman named, across a network with no centre of the kind a Polish-born American designed, with the checking left to the ends as a Frenchman's network first did it. It passed through a post office that never opens letters, found its way with a phone book nobody holds whole, and reached an Englishman's three inventions, given to everyone by a laboratory near Geneva. Some of those pieces earned their makers money. CSIRO collected its licence fees, and the companies that lay cables and run data centres charge for them. But the parts that let all of it speak one language, the envelope, the post office rules, the rules of the phone book and the web itself, are published for anyone to use, and the web's standards are written so that anyone can build on them without paying royalties.

So when someone asks who invented the internet, the reply is a list rather than a name, and most of the people on it handed their piece on without sending an invoice. If the next page loads slowly in a block of flats, your phone is most likely waiting its turn in a crowded room. The rest of the route is holding: a few dozen people's ideas, most of them given away decades ago, and somewhere under the Baltic a cable whose strands of glass are no thicker than a hair, still in one piece.