by Andrew Blum
Boelter Hall was where the Internet had once been fully containable, in stark contrast with its current sprawl. And Kleinrock was still there, embodying that history in his memory. But I could have reached him on the phone, we could have video-chatted. But I had cast my net in the waters of experience, having chosen (for example) to ignore the photograph of room 3420 that shows up in a Google image search, and instead come see it for myself. That afternoon, when I had arrived early for my appointment with Kleinrock, I sat on a ledge outside Boelter Hall eating a bag of chips, fiddling with my cell phone. My wife had just emailed a video of our baby taking her first crawling steps, a video that loaded vividly on the small screen and pulled me back, in my head, to New York. I had come to visit the first node of the Internet, but one of its most recent nodes—the one I carried in my pocket—had distracted me. If the Internet was a fluid new world distinct from the old physical world, Boelter Hall struck me as a place where the two met, forming an unusually visible seam. Except the essence I sought was diluted by the evolution of the thing that it created. Here was the shiny new device, connected everywhere; there was the ancient machine in the wood case, smelling of mildew. What, really, was the difference? The IMP here was the real thing: not a replica or a model—or a digital image. That’s why I was here: to hear the details from Kleinrock himself and note the color of the walls, but also to thumb my nose at the immediate reproducibility of everything else. The place itself couldn’t be blogged and reblogged—and I confess that I was a little drunk on the irony of that. In his 1936 essay, “The Work of Art in the Age of Mechanical Reproduction,” Walter Benjamin describes the fading importance of an object’s “aura,” its unique essence; and here I was in search of the aura of the thing that threatened to destroy the idea of “aura” once and for all.
I asked Kleinrock about this: Why isn’t essence a word we typically talk about in the context of the Internet itself? It’s more often the opposite that thrills us: the network’s ease at instant reproduction, its ability to make things “viral,” with the consequence of threatening not only the aura but also our desire for it—leading us to watch a concert through a smartphone screen. “For the same reason people don’t know when it was created or where it started, or what the first message was,” he said. “It’s an interesting psychological and sociological commentary that people are not curious about it. It’s like oxygen. People don’t ask where oxygen comes from. I think the students today miss a lot because they don’t take things apart. You can’t take this apart”—he rapped his new laptop. “Where’s the physical experience? Unfortunately it’s gone. They have no idea how this thing works. When I was a kid building radios I knew what I was dealing with; I knew how things worked and why they worked that way.” The lab then in progress in room 3420 was an exception, the one time the computer science students got their hands dirty.
I asked Kleinrock about some of the mementos lying around his office. From a small gray archival box resting on top of a file cabinet he pulled out the original log recording the moment when UCLA’s IMP first connected with IMP #2, installed at Stanford Research Institute, late on the Wednesday evening of October 29, 1969. The notebook was tan, with “IMP LOG” written in sloppy marker on its cover. You can see it on Kleinrock’s website, of course. “That’s the most precious document on the Internet,” he said. “There is someone putting together these archives now, and they shoot me every time I touch this. They’re the ones who gave me this box.” He opened it up and began to read the entries:
SRI called, tried debug test, but it didn’t work.
Dan pushed some buttons.
“The important one is here—October twenty-ninth. I shouldn’t be touching these pages, but I can’t resist! There it is.” In blue ballpoint pen, the words running over two lines and beside the time code 22:30, was written:
Talked to SRI host to host.
It is the sole documentary evidence of the ARPANET’s first successful transmission between sites—the moment of the Internet’s first breath. I nervously kept my hands in my lap. “If anybody comes to steal it, here it is!” Kleinrock said. “There’s also a copy of my dissertation.”
Then he turned nostalgic. “In those early days, none of us had any idea what it would become. I had a vision, and I got a lot of it right. But what I missed was the social side—that my ninety-nine-year-old mother would be on the Internet when she was alive. That part eluded me. I thought it was going to be computers talking to computers or people talking to computers. That’s not what it’s about. It’s about you and me talking.”
I reminded him that we had closed up the IMP to let the smell “cook up,” and we again crossed the hall to pay our respects. Kleinrock opened the cabinet. “Here,” he said. “Yeah. Mmmm. Get your nose down there.” I leaned in, as if toward a flower. “Smell that? There are components here, there’s rubber. This is the stuff when I was a kid I used to cannibalize old radios, with vacuum tubes, and I would smell the solder a lot, the resin.” I recalled the electronics class I took as a third grader, after school. We made LEDs that blinked in a pattern. I spend my days connected to electronic machines but I’d hardly smelled it since.
“You can’t record that,” Kleinrock said. “Yet. One day you will.”
The Internet’s adolescence was protracted. From the ARPANET’s birth at UCLA in 1969 until the mid-1990s, the network of networks crept slowly outward from universities and military bases to computer companies, law firms, and banks, long before it found its way to the rest of us. But in those long early years there really wasn’t much of it to speak of. For a quarter century, Kleinrock and his colleagues were like explorers, staking the flag of the nascent Internet on a series of far-flung colonies, connected only tenuously with one another, and often not at all with their immediate surroundings. The Internet was thin on the ground.
The early maps of the ARPANET frequently published by Bolt, Beranek and Newman show just how thin. They look like constellation charts. In each version, an outline drawing of the United States is overlaid with black circles indicating each IMP, linked by razor-straight lines. The ARPANET began life as the Little Dipper, scooping up a piece of California, its handle in Utah. By the summer of 1970, it had expanded east across the country to add MIT, Harvard, and Bolt’s Cambridge offices. Washington didn’t appear until the following fall. By September 1973, the ARPANET went international, with the establishment of a satellite link to University College London. By the end of the decade, the network’s geography was fully entrenched around four regions: Silicon Valley, Los Angeles, Boston, and Washington. New York City hardly appeared at all, with only NYU winning a colony. Only a few scattered nodes dotted the middle of the United States. True to its philosophical roots as a doomsday-era communications system, the ARPANET was strikingly deurbanized and decentralized. It had no special places, no monuments. Physically speaking, there were IMPs like the one down the hall from Kleinrock’s office, linked by always-on phone connections provided under special terms by AT&T. It existed in spare classrooms of university computer science departments, within outbuildings on military bases, and across the copper lines and microwave links of the existing telephone network. The ARPANET wasn’t even a cloud. It was a series of isolated outposts strung together by narrow roads, like a latter-day Pony Express.
No doubt there was serious research going on, but the ARPANET’s use as a communications tool retained an air of novelty. In September 1973, a conference at Sussex University, in Brighton, England, brought together computer scientists from around the world who were each developing their own government-sponsored computer networks. Since the ARPANET was the largest, a special demonstration link was established back to the United States. It hadn’t been an easy thing to arrange. A telephone line had to be activated between one of the ARPANET nodes in Virginia and a nearby satellite antenna. From there, the signal was bounced off an orbiting satellite to another earth station in Goonhilly Downs in Cornwall, then onward through a telephone link to London,
and finally into Brighton. It was less a technological marvel than what engineers like to call a “kludge,” a temporary and tenuous link across the ocean.
But history remembers it for more prosaic reasons. In a story that has become legend, when Kleinrock arrived home in Los Angeles from the conference, he realized he’d forgotten his electric razor in the Sussex dormitory bathroom. Logging into the ARPANET from his UCLA computer terminal, he entered the command WHERE ROBERTS, which told him if his friend Larry Roberts—a well-known workaholic and insomniac—was logged in as well. Sure enough he was, wide awake at 3:00 A.M. Using a rudimentary chat program—“clickety clickety clack,” as Kleinrock describes it—the two friends made arrangements to send home the razor. That kind of communication was “a bit like being a stowaway on an aircraft carrier,” as the historians Katie Hafner and Matthew Lyon describe it.
The 1970s ARPANET was US government property, linking defense researchers either within the military itself or at ARPA-funded university departments. But socially the ARPANET was a small town. The 1980 edition of the ARPANET directory is a canary-colored perfect-bound book, about the thickness of a fall fashion magazine. It lists the five-thousand-odd names of everyone on the ARPANET, with their postal addresses, the lettered code for their nodes, and their email addresses—absent the “.com” or “.edu,” which wouldn’t be invented for another few years. Kleinrock is in there, of course, with the same office address and phone number as today (although his area code, zip code, and email address have all changed). Sharing the page with him are computer scientists at MIT, University College London, and the University of Pennsylvania; a commander at the Army Communications Research and Development Command at Fort Monmouth, New Jersey; and the chief of the Strategic Studies Programming Division at Offutt Air Force Base in Nebraska—famous as the manufacturing site of the Enola Gay, the primary Cold War–era nuclear command center, and the place at which President Bush temporarily sought refuge on 9/11.
The ARPANET was like that: an accidental meeting place for academics and high-tech soldiers, brought together under the umbrella of computer networking. Inside the front flap of the directory is a logical map of the ARPANET, with each node labeled in tiny print and connected together with thick and thin straight lines, like an elaborate and convoluted flowchart. Every computer on the ARPANET fits easily on the page. But that intimacy wouldn’t last.
By the early 1980s, the big computer companies—like IBM, XEROX, or Digital Equipment Corporation—and large government agencies—like NASA and the Department of Energy—were running their own independent computer networks, each with its own acronym. High-energy physicists had HEPnet. Space physicists had SPAN. Magnetic fusion researchers had MFENET. A handful of European networks had also emerged, including EUnet and EARN (the European Academic Research Network). And there were a growing number of regional academic networks, named like the twelve sons of Mr. and Mrs. Net: BARRNet, MIDnet, Westnet, NorthWestNet, SESQUINet.
The trouble was, all those networks weren’t connected. While stretching nationwide and occasionally across the ocean, they operated in effect as private highways overlaid on the public telephone system. They overlapped geographically, sometimes serving the same university campuses. And they even might have overlapped physically, sharing the very same long-distance telephone cables. But in networking terms they were “logically” distinct. They were disconnected—as separate as the sun and the moon.
That remained the case until New Year’s Day 1983 when, in a transition years in the planning, all the host computers on the ARPANET adopted the electronic rules that remain the basic building block of the Internet. In technical terms, they switched their communications protocol, or language, from NCP, or “Network Control Protocol,” to TCP/IP, or “Transmission Control Protocol/Internet Protocol.” This was the moment in the Internet’s history when the child became father to the man. The changeover, led by the engineers at Bolt, Beranek and Newman, kept dozens of system administrators tied to their desks on New Year’s Eve, struggling to make the deadline—leading one to commemorate the ordeal by making I SURVIVED THE TCP/IP TRANSITION buttons. Any node that did not comply was cut off until it did. But once the dust had settled several months later, the result was the computing equivalent of a single international language. TCP/IP went from a dominant dialect to an official lingua franca.
As the historian Janet Abbate notes, the changeover marked not just an administrative shift but a crucial conceptual one: “It was no longer enough to think about how a set of computers could be connected; network builders now also had to consider how different networks could interact.” The ARPANET was no longer a walled garden with an official government directory of participants, but rather had become just one network among many, linked together into an “Internetwork.”
The New Year’s 1983 standardization of TCP/IP permanently fixed the Internet’s distributed structure, ensuring to this day its lack of central control. Each network acts independently, or “autonomously,” because TCP/IP gives it the vocabulary to interact. As the author and Columbia law school professor Tim Wu points out, this is the founding ideology of the Internet, and it has clear similarities with other decentralized systems—most notably the federal system of the United States. Because the early Internet ran on the existing wires of the telephone network, its founders were forced “to invent a protocol that took account of the existence of many networks, over which they had limited power,” Wu writes. It was “a system of tolerated difference—a system that recognized and accepted the autonomy of the network’s members.”
But while this autonomy came about because of the infrastructure the Internet was given, it soon became the crucial force shaping the infrastructure the Internet made. Winston Churchill said about architecture that “we shape our buildings, and afterwards our buildings shape us,” and the same is true of the Internet. With TCP/IP in place and new autonomous networks popping up with increasing frequency, the Internet grew physically, but haphazardly. It took shape in ad hoc ways, like a city, with a loose structure giving way to spontaneous, organic growth. The Internet’s geography and shape weren’t drawn up in some central AT&T engineering office—as the telephone system was—but rather arose out of the independent actions of first hundreds, and later thousands, of networks.
With TCP/IP in place, the Internet—more or less as we know it today—had arrived, and a remarkable period of growth began. In 1982, there were only 15 networks, or “autonomous systems,” on the Internet, meaning they communicated with TCP/IP; by 1986 there were more than 400. (In 2011, there were more than 35,000.) The numbers of computers on those networks ballooned even faster. In the fall of 1985, there were 2,000 computers with access to the Internet; by the end of 1987 there were 30,000, and by the end of 1989 there were 159,000. (In 2011, there were 2 billion Internet users, with their hands on even more devices.) The Internet, which had for nearly twenty years been a college town called the ARPANET, had begun to feel more like a metropolis. If before you could imagine each router as a cloister on a quiet mountaintop, the incredible growth in the number of machines meant that those routers were now piling up near one another, forming villages. Some of those villages were even beginning to reveal the vague promise of a skyline. For me, it’s the most exciting moment of this early history: the Internet was becoming a place.
Toward the end of the 1980s a handful of companies began to build their own long-distance data highways, or “backbones,” and city streets, or “metropolitan” networks. But strike from your mind any images of bulldozers steaming across the Pennsylvania countryside laying cable—although those would come soon enough. These early long-distance and local networks still worked across the existing phone lines, with specialized equipment installed on either end. By the early 1990s, the trickle became a wave, as companies like MCI, PSI, UUNet, MFS, and Sprint attracted increasing investment dollars—and used them to dig their own trenches and fill them with the new technology of fiber optics, which had been commerc
ialized in the 1980s. The network of networks was accumulating an infrastructure of its own. It began to colonize key places around the world—indeed, the places where it still predominantly exists: suburban Virginia and Silicon Valley, California; London’s Docklands district; Amsterdam, Frankfurt, and downtown Tokyo’s Otemachi district. The Internet had propagated to the point of becoming visible to the naked eye, becoming a real landscape all its own. What for the first twenty years of the Internet’s existence were easy to dismiss as in-between spaces—telecom closets and spare classrooms—now had character. By the mid-1990s the wave of construction became a torrent, and “broadband” became one of the most infamous bubbles in American economic history. Yet it was that spending, as overheated and economically destructive as it was, that built the Internet we use today.
In 1994, I was finishing high school, logging long hours on the family Macintosh, endlessly tying up the phone exploring the message boards and chat rooms of America Online. Then sometime that winter my father brought home a small 3.5-inch disk loaded with a new program called “Mosaic”—the first web browser. On a sunny weekend morning, sitting at the dining room table, my physics homework pushed to the side, a long telephone cord strung across the room, we listened as the screeching tones of the modem signaled a connection with a distant computer. My mother looked disapprovingly over the top of her newspaper. On the screen, rather than the America Online menu, with its short list of choices, there was a blinking cursor inside an empty “address bar”—the ur-starting point for all our digital journeys.