Internet History: ARPANET - The Origin

Original author: Creatures of Thought
  • Transfer

By the mid-1960s, the first time-sharing computing systems generally repeated the early history of the first telephone switches. Entrepreneurs created these switches to allow subscribers to use the services of a taxi, doctor, or fire brigade. However, subscribers soon discovered that local switches are just as suitable for communicating and socializing with each other. Likewise, time-sharing systems, first created so that users could “call out” computing power, soon turned into utility switches with an integrated messaging system. In the next decade, computers will go through the next stage in the history of the telephone - the emergence of the interconnection of switches, forming regional and long distance networks.


The first attempt to integrate several computers into a larger coherent whole was the SAGE interactive computer network project, American air defense system. Since each of the 23 SAGE control centers covered a specific geographic area, a mechanism was needed to transmit radar tracks from one center to another, in cases when a foreign aircraft crossed the border between these areas. SAGE developers called this task “cross-telling,” and solved it by creating data lines based on dedicated AT&T telephone lines stretched between all neighboring control centers. Ronald Enticnap, who was part of a small delegation of the Royal Armed Forces sent to SAGE, led the development and implementation of this subsystem. Unfortunately, I did not find a detailed description of the “inter-colloquial” system, but, obviously, the computer in each of the control centers determined the moment when the track on the radar passed to another sector,

The SAGE system needed to translate digital data into an analog signal on the telephone line (and then back to the receiving station), and therefore AT&T was given the opportunity to develop the Bell 101 modem (or dataset, as it was called at first), capable of transmitting a modest 110 bit per second. This device was later called a modem , for its ability to modulate an analog telephone signal using a set of outgoing digital data, and demodulate bits from an incoming wave.

Bell 101 dataset

Thus, SAGE laid an important technical basis for later computer networks. However, the first computer network, whose legacy was long and influential enough, was the network with the name, known today: ARPANET. Unlike SAGE, it combined a diverse set of computers, both with user time sharing and with batch processing of data, each of which had its own special set of programs. The network was conceived as universal in scale and operation, and had to satisfy any user needs. The project was funded by the Information Processing Techniques Office (IPTO), led by director Robert Taylor, a department of computer research at ARPA. But the very idea of ​​such a network came up with the first director of this department, Joseph Karl Robnett Liklider.


As we learned earlier , Liklider, or “Lik” for his colleagues, was a psychologist by training. However, when he worked with radar systems in the Lincoln lab in the late 1950s, he was fascinated by interactive computers. This passion led him to finance some of the first experiments on computers with time-shared access, when in 1962 he became director of the newly formed IPTO.

By then, he was already dreaming of the possibility of linking isolated interactive computers into a larger superstructure. In his 1960 work on the "symbiosis of man and computer," he wrote:
It seems reasonable to imagine a “thought center” that can incorporate the functions of modern libraries and the alleged breakthroughs in the field of storage and retrieval of information, as well as the symbiotic functions described above in this paper. This picture is easily scaled into a network of such centers, united by broadband lines, and accessible to individual users through leased telephone lines.

Just as the TX-2 sparked Lick’s passion for interactive computers, SAGE might have prompted him to imagine how various interactive computing centers could be connected together and provide something like a telephone network for intelligent services. Wherever this idea came from, Lick began to spread it over the research community created by him in the IPTO, and the most famous of these messages was a memo dated April 23, 1963, addressed to “Members and departments of the intergalactic computer network,” that is, various researchers who received funding from IPTO for computer access with time sharing and other computing projects.

The note looks messy and chaotic, clearly dictated on the fly and not edited. Therefore, in order to understand what exactly Lik wanted to say about computer networks, one has to think a little. However, some points are immediately highlighted. First, Lick said that the “various projects” funded by the IPTO actually belong to “one area.” After that, he discusses the need to allocate money and projects to maximize the benefits of this enterprise, because among the network of researchers “to achieve progress, every active researcher needs a software base and equipment more complex and comprehensive than he can create in a reasonable amount of time.” Leek concludes that achieving this global effectiveness requires some personal concessions and sacrifices.

Then he begins to discuss in detail computer (rather than social) networks. He writes about the need for some kind of network management language (what will later be called the protocol) and his desire to someday see the IPTO computer network, consisting of “at least four large computers, possibly six to eight small computers, and a large assortment disk drives and magnetic tape - not to mention remote consoles and teletype stations. ” Finally, he outlines on several pages a concrete example of how interaction with such a computer network may develop in the future. Lik imagines a situation in which he analyzes some experimental data. “The problem,” he writes, “is that I don’t have a decent program for building charts. Is there a suitable program somewhere in the system? Using the doctrine of network dominance, I first interview the local computer, and then other centers. Suppose I work in the SDC, and that I find a seemingly suitable program on disk in Berkeley. " He asks the network for the execution of this program, suggesting that “with a complex network management system, I don’t have to decide whether to transfer data so that the programs process it somewhere else, or download the programs to myself and run them to work on my data.”

Together, these fragments of ideas open up a larger scheme conceived by Liklider: first, to divide certain specialties and fields of knowledge between researchers receiving funding from IPTO, and then build a physical network from IPTO computers on the basis of this social community. This physical manifestation of the “common cause” of the IPTO will allow researchers to share knowledge and gain the benefits of specialized hardware and software at each workstation. In this way, IPTO will be able to avoid wasteful duplication, while enhancing the capabilities of every dollar of funding, giving each researcher of all IPTO projects access to a full range of computing capabilities.

This idea of ​​sharing resources between members of the research community through a communications network has thrown seeds into the IPTO that sprouted several years later in the form of creating ARPANET.

Despite its military background, the ARPANET, which appeared at the Pentagon, had no military justification. It is sometimes said that this network was developed as a military communications network capable of surviving a nuclear attack. As we will see later, there is an indirect connection between ARPANET and an earlier project with this purpose, and ARPA executives periodically talked about “fortified systems” to justify the existence of their network before Congress or the Secretary of Defense. But in fact, IPTO created ARPANET purely for its internal needs, to support a community of researchers - most of which could not justify their activity by working for defense purposes.

Meanwhile, at the time of the release of his famous memo, Lyclider had already begun to plan the germ of his intergalactic network, whose director would be Leonard Kleinrock of the University of California at Los Angeles (UCLA).

Console for SAGE model OA-1008, complete with a light gun (at the end of the wire, under a transparent plastic cover), a lighter and an ashtray.


Kleinrock was the son of East European working-class immigrants, and grew up in Manhattan in the shadow of the im. George Washington [ connects upstate Manhattan in New York City and Fort Lee in Bergen County in New Jersey.] ]. While studying at school, in the evenings, he took additional lessons in electrical engineering at New York City College. Hearing of the opportunity to study at MIT followed by a semester of full-time work at Lincoln’s lab, he happily grabbed onto it.

The laboratory was created to serve the needs of SAGE, but has since spread to many other research projects, often only indirectly related to air defense, if at all related to defense. Among them was the Barnstable Research project, a concept proposed by the Air Force to create an orbital belt of metal strips (like dipole reflectors ) that could be used as a global communications system. Kleinrock conquered the authority of Claude Shannonfrom MIT, so he decided to concentrate on the theory of communication networks. Barnstable's study gave Kleinrock the first opportunity to apply information theory and queuing theory to a data network, and he expanded this analysis to a whole dissertation on messaging networks, combining mathematical analysis with experimental data collected from simulations running on TX-2 computers in laboratories Lincoln. Among Kleinrock’s close associates in the lab, who used computers for the time sharing system, were Lawrence Roberts and Ivan Sutherland , whom we will meet a little later.

By 1963, Kleinrock accepted a job offer at UCLA, and Liklider saw it as an opportunity. In front of him was an expert on data networks working near three local computer centers: the main computer center, the health computer center, the Western data center (a cooperative of thirty institutes that had shared access to an IBM computer). Moreover, six institutes from the Western data center had a remote modem connection to the computer, and the IPTO sponsored System Development Corporation (SDC) computer was just a few kilometers from Santa Monica. IPTO ordered UCLA to bring these four centers together as the first experiment in building a computer network. Later, according to the plan, communication with Berkeley could study the problems inherent in long-distance data transmission.

Despite the promising situation, the project failed and the network was never built. The directors of the various UCLA centers did not trust each other, and did not believe in this project, which is why they refused to give up control of computing resources to each other's users. IPTO had practically no leverage to influence this situation, since not one of the data centers received money from ARPA. This political issue points to one of the main issues in the history of the Internet. If you convince different participants that the organization of communication between them and cooperation plays into the hands of all parties, it is very difficult, how did the Internet come about? In the following articles we will return to these questions more than once.

IPTO’s second attempt to build a network proved to be more successful, perhaps because it was much less extensive — it was a simple experimental test. And in 1965, a psychologist and student of Liklider by the name of Tom Marill left the Lincoln laboratory to try to make money on the hype about interactive computers, starting his own business to provide shared access. However, without gaining a sufficient number of paid customers, he began to look for other sources of income, and in the end offered IPTO to hire him to conduct research on computer networks. IPTO's new director, Ivan Sutherland, decided to partner with a large and reputable company as ballast, and subcontracted Maryll through Lincoln’s lab.

Roberts, as a student at MIT, was adept at working with the TX-0 computer built by the Lincoln lab. He sat fascinated for hours in front of the glowing console screen, and eventually wrote a program that (badly) recognized handwritten characters using neural networks. Like Kleinrock, he eventually began working for the laboratory for graduate studies, solving tasks related to computer graphics and computer vision, for example, face recognition and the generation of three-dimensional images, on a larger and more powerful TX-2.

For most of 1964, Roberts focused mainly on imaging. And then he met with Lick. In November of that year, he attended a conference on the future of computers, sponsored by the Air Force, and held at a hot spring sanatorium in Homestead, West Virginia. There, until late at night, he talked with other participants in the conference, and for the first time he heard Lick presenting his idea of ​​an intergalactic network. Roberts something stirred in his head - he did a great job with processing computer graphics, but, in fact, was limited to one unique computer TX-2. Even if he could share his software, no one else could use it, since no one had the equivalent equipment to run it. The only way to expand the influence of his work for him was to talk about them in scientific papers, in the hope that someone can reproduce them somewhere else. He decided that Lick was right - the network was exactly the next step that needed to be done to accelerate research in the field of computer technology.

And Roberts ended up working with Marill, trying to connect the TX-2 from Lincoln’s lab via a telephone line across the country to an SDC computer in Santa Monica, California. In a pilot project, as if copied from Lick's memo on the “intergalactic network”, they planned to get the TX-2 to pause work in the middle of the calculations, use the automatic dialer to call SDC Q-32, run the matrix multiplication program on that computer, and then continue the initial calculations using his answer.

In addition to the meaningful use of expensive and advanced technology in order to transmit the results of a simple mathematical operation across the entire continent, it is worth noting the terribly low speed of this process due to the use of the telephone network. To make a call, it was necessary to set up a dedicated connection between the caller and the called person, which usually went through several different telephone exchanges. In 1965, almost all of them were electromechanical (this year AT&T launched the first all-electric station in Sakasuna, New Jersey). Magnets moved metal bars from one place to another to provide contact in each of the nodes. The whole process took several seconds, during which the TX-2 just had to sit and wait. In addition, the lines perfectly adapted for conversations, were too noisy to transmit individual bits, and provided very low bandwidth (a couple of hundred bits per second). A truly effective intergalactic interactive network required a different approach.

The Merrill-Roberts experiment did not demonstrate the practicality or usefulness of the long-distance network, showing only its theoretical performance. But that was enough.


In mid-1966, Robert Taylor became the new, third director of the IPTO, following Ivan Sutherland. He was a student of Liklider, also a psychologist, and joined the IPTO because he had previously administered computer science research at NASA. Apparently, almost immediately upon arrival, Taylor decided that it was time to realize the dream of an intergalactic network; it was he who launched the project that spawned ARPANET.

Money from ARPA was still flowing, so Taylor had no problem getting extra funding from his boss, Charles Herzfeld. However, this decision had a significant risk of failure. Besides the fact that in 1965 there were quite few lines connecting the opposite ends of the country, no one had previously tried to do anything similar to ARPANET. You can recall other early experiments in the creation of computer networks. For example, Princeton and Carnegie Mallon raised the grid from shared-access computers in the late 1960s with IBM. The main difference of this project was its homogeneity - it used exactly the same hardware and software computers.

ARPANET, on the other hand, would have to deal with diversity. By the mid-1960s, IPTO funded more than ten organizations, each of which had a computer, all of which worked on different equipment and with different software. The ability to share software was rarely even among different models of the same manufacturer - they decided to do this only with the latest line of IBM System / 360.

The variety of systems was a risk that added both considerable technical complexity to the development of the network and the possibility of sharing resources in the style of Liklider. For example, at the University of Illinois at that time, a massive supercomputer ILLIAC IV was built with ARPA money. It seemed unlikely to Taylor that local users from Urbana-Campaign could fully utilize the resources of this huge machine. Even systems on a much more modest scale — TX-2 at the Lincoln lab and Sigma-7 at UCLA — usually could not share software with each other due to fundamental incompatibilities. The ability to overcome these limitations by gaining direct access to the software of one node while in another was attractive.

In a paper describing this network experiment, Marill and Roberts suggested that such a resource exchange would lead to something like a Ricardian comparative advantage for computing nodes:
Networking can lead to a certain specialization of collaborating nodes. If a certain node X, due to the presence of special software or equipment, for example, is particularly successful in inverting matrices, it can be expected that users of other network nodes will use this feature by inverting their matrices on the node X, instead of doing this on their own home computers.

Taylor had another motivation for implementing a resource sharing network. Buying a new computer for every new IPTO node that had all the features that the researchers could ever need on this node was expensive, and as the new nodes were added to the IPTO portfolio, the budget expanded dangerously. By linking all IPTO-funded systems into one network, it will be possible to provide newer grant recipients with more modest computers, or even not buy them at all. They could use the computer power they needed on remote sites with an excess of resources, and the entire network would work as a public reservoir of software and hardware.

After launching the project and securing its financing, Taylor’s last significant contribution to ARPANET was the choice of the person who will directly engage in the development of the system and see to it that it is implemented. The obvious choice was Roberts. His engineering skills were not in doubt, he was already a respected member of the IPTO research community, and he was one of the few people who had real experience in designing and building computer networks operating over long distances. So in the fall of 1966, Taylor called Roberts and asked him to come from Massachusetts to work on the ARPA in Washington.

But it was difficult to seduce him. Many IPTO academic leaders were skeptical of Robert Taylor's reign, considering him lightweight. Yes, Liklider was also a psychologist, did not have an engineering degree, but at least he had a doctorate, and certain merits as one of the founding fathers of interactive computers. Taylor was an unknown person with a master's degree. How can he manage the complex technical work in the IPTO community? Roberts was also among these skeptics.

But the combination of carrots and sticks did the trick (most sources point to the predominance of whips in the absence of gingerbread). On the one hand, Taylor put some pressure on the head of Roberts in the Lincoln laboratory, reminding him that most of the funding for the laboratory now comes from ARPA, and that therefore he should convince Roberts of the benefits of this proposal. On the other hand, Taylor proposed to Roberts the recently established title of “senior scientist,” who will report directly to ARPA's deputy director through Taylor’s head, and will also become Taylor’s successor as director. Under these conditions, Roberts agreed to take up the ARPANET project. The time has come to turn the idea of ​​sharing resources into reality.

What else to read

  • Janet Abbate, Inventing the Internet (1999)
  • Katie Hafner and Matthew Lyon, Where Wizards Stay Up Late (1996)
  • Arthur Norberg and Julie O'Neill, Transforming Computer Technology: Information Processing for the Pentagon, 1962-1986 (1996)
  • M. Mitchell Waldrop, The Dream Machine: JCR Licklider and the Revolution That Made Computing Personal (2001)

Also popular now: