BY Steve Crocker |
Steve Crocker is chairman of the Internet Corporation for Assigned Names and Numbers. In the late 1960s, he was a UCLA graduate student who helped create the ARPANET, a precursor to the Internet.
During the early and mid 1960s – about a half a century ago – computers were physically very big. Only big companies, universities and governments had computers, storing them in special air-conditioned rooms. And they were expensive. On IBM’s flagship mainframe computer, the IBM 7094, the memory unit – what you would now call the RAM – held one megabit, i.e. about 128KB, and cost about one million dollars in the early 1960s. That’s about five million dollars in today’s terms. Each computer was also the center of its own universe. There were a few experiments and special projects to connect computers together, but nothing common or easy.
Here’s how the government changed all of that. In 1958, the Defense Advanced Research Projects Agency (DARPA) was created in response to the Russians putting up Sputnik, the first satellite. American officials were chagrined that the Russians had beaten them into space, and wanted to make sure it stayed ahead in research in the future. One portion of DARPA was the Information Processing Techniques Office (IPTO), which funded research at major universities around the U.S. focusing on the frontiers of computer science in that period – man-computer interaction, artificial intelligence, large-scale computers, advanced graphics, all much of what became today’s common technology.
By the mid-1960s IPTO was funding a few dozen projects at UCLA, MIT, Harvard, Carnegie-Mellon, Stanford and similar universities and research laboratories. An important aspect of these efforts is that each one was unique. The researchers were free to choose what equipment to use and how to run their projects, so there wasn’t any uniformity across these projects.
From their central vantage point in DARPA, key people could see the value of connecting these laboratories together. The directors of DARPA during the 1965 to 1975 time period, Charles Herzfeld, Eb Rechtin and Steve Lukasik , all understood the pressing need within the U.S. military to have technology for connecting and utilizing the ever more complex computer systems the military was building. IPTO’s first three directors, JCR Licklider, Ivan Sutherland, and Bob Taylor, all initiated efforts to connect computers together.
The first few attempts didn’t succeed very well. Finally, Bob Taylor, after getting advice from many of the researchers in the field, proposed a general-purpose network that would span the country and connect all of these laboratories. The goal included connecting the machines and the people. Thus, he envisioned not just computers working with each other in support of a single project, i.e. “resource sharing,” but also people working with each other and forming new working relationships.
That project was known as the ARPANET, and it transformed the world. The planning for the ARPANET took place during 1967-68, and the construction took place starting in 1969. I had the privilege of being one of the graduate students at the UCLA research project that became the first node on the ARPANET. We connected the first computer to the net. More important, in cooperation with our colleagues in the other laboratories, we tackled the problem of how to connect very different kinds of computers and have them talk to each other in a useful way. These rules are the “protocols” and we realized right from the start that there were going to be a lot of protocols with more to be developed every year.
The ARPANET was a big success as a research project. Among its several transformative innovations was the first large-scale use of packet switching. Packets are short messages – a thousand bits or less in the Arpanet; closer to ten thousand bits in today’s networks – that are individually routed through the network. Breaking up long streams into packets permits far greater flexibility and efficiency in the use of the expensive circuits.
Within a few years, all of the DARPA computer science research sites were connected to each other, and it was starting to make a big difference in the quality and nature of the research across these laboratories. People began to work together in ways that had been impossible before. Just imagine a world without email, messaging, file sharing, etc.
From One Network to The Internet
The ARPANET was also a big success in stimulating other network projects, and here’s where things get interesting. The ARPANET was a single network, but it was quite clear there were going to be other packet-switched networks. Even then, it was equally clear they had to be interconnected. There were at least three different reasons for other networks.
First, the ARPANET was built using fixed telephone lines between the nodes. Each node had an “IMP” – the name stood for Interface Message Processor – and it was the first version of what we now call a router. The IMPs used packet switching to move information from one computer to another, thereby making very efficient use of the very expensive telephone lines. The notion of packet switching is very powerful and works well in radio environments too. IPTO capitalized on the success of the ARPANET and initiated work on a packet-radio network, somewhat similar to today’s CDMA wireless technology, and on a packet satellite network. Connecting both the packet radio network and the satellite network to the ARPANET was the starting point for today’s Internet as a global collection of interconnected networks.
A second impetus toward multiple networks was administrative. Within the U.S. government, DARPA’s mandate was research of relevance to the military. At least two other parts of the government, NASA and the Department of Energy, also funded work at several laboratories, and they too started to build networks. Again, it was evident these all would benefit from being interconnected, but it was also clear the interconnection had to treat all of the networks with equal status, i.e. none subservient to any other. Even though all were funded and run by the U.S. Government, each agency required complete control over the administration and operation of its own network.
A third impetus came from outside the United States. Other countries, seeing the success of the ARPANET, started their own networks. Canada started CANUNET to connect its universities, France started Cyclades, and the British had already been working on packet switching at the National Physical Laboratory. Again, it was evident users of these networks would eventually need to be to interact with users on other networks.
Two of the key people who create this interconnection were Vint Cerf and Bob Kahn. Vint and I had been close friends since high school, and we wound up sharing an office at UCLA when the ARPANET was started. Bob was part of the group at Bolt, Beranek and Newman (BB&N) in Cambridge, MA that designed and built the IMPs for the ARPANET. Bob and I each joined IPTO in Washington in the early 1970s. Vint finished his PhD and went to Stanford to teach for a while and later joined DARPA. Meanwhile, the ARPANET had not only brought people together to work on the projects I mentioned earlier – artificial intelligence, graphics, etc. – it also created a whole new field of research devoted to the science and engineering of networks.
What started out as a loose set of interactions among the graduate students and professors in the laboratories soon evolved into a vibrant community of people specializing in this new area of research of network design and operation. Research reports and other publications, regular meetings, and many, many experiments resulted. And, of course, a central goal during this period was how to interconnect the many disparate networks that were being built and to do so in a way that permitted independent operation but effective communication.
Bob and Vint wound up collaborating closely, both with each other and with the rest of the community, and evolved the TCP and IP protocols. These built on the experience from the ARPANET and the earlier NCP (Network Control Protocol) that I had spearheaded. Their design was lean, which was the key. In keeping the design lean and simple, they left room for each network to be built and operated in its own way. It also made it possible for a wide range of the applications to be built on top of these very basic protocols.
As the initial research networks flourished, there was pressure to connect the rest of the universities and colleges in the U.S. Under the leadership of Dave Farber and Larry Landweber, the National Science Foundation—another key government agency--played a major role in creating CSNET and then the NSFNET, a national network that connected several regional level networks to link several thousand research universities together with the rest of the Internet. It also invested in international connections to link other academic networks into the global and growing Internet.
All of these developments have been documented in depth, both in the technical and popular literature. Bob and Vint, along with many others including Bob Taylor and Larry Roberts, the next IPTO director and the one who really brought the ARPANET to life, Bob Metcalfe, the inventor of the Ethernet, Len Kleinrock for his contributions to packet switching, Doug Engelbart for his work on human computer interaction including the invention of the mouse, Dave Farber and Larry Landweber for their broad expansion of the Internet within the academic community, and, many years later, Tim Berners-Lee for his work on the invention of the World Wide Web, all received a lot of attention, awards, and, I’m sure, satisfaction. Even my brother David and I also got awards, he for his work on standardizing email and I for my early work on the general protocol architecture, development of the community that continued to work on networks and, most peculiarly, for my quixotic creation of what I thought was going to be a temporary series of notes, the Request for Comments.
You Didn’t Build That
Could the Internet have been created by private industry? Without government’s help as funder and convenor? I don’t think so. Here’s why.
When the ARPANET was created, each of the research laboratories around the country had different equipment. That meant we had to figure out how to connect machines that were manufactured by different companies. The first four computers on the ARPANET were the Xerox Data Systems Sigma 7 at UCLA, the Scientific Data Systems (later acquired by Xerox Data Systems) SDS 940 at SRI International, the IBM 360/75 at the University of California, Santa Barbara, and the Digital Equipment PDP-10 at the University of Utah. In order to connect these computers, there had to be some sort of common standard that was not controlled by any single company. This applied to every level of the system, from the hardware interface to the IMP to the basic protocols that moved messages from one computer to another, to the higher level applications such as email and, later, the world wide web. The open architecture of the Internet, with defined interfaces and open standards that were all freely available, initially made it possible for any person and any company to participate. This was a fundamental principle of the early work. This was also a hallmark of the government research effort, and would not have come about if the Internet had been created by industry.
On the other hand, government funding can only play a small role in the development of something as big as the Internet. With the initial infrastructure in place, it was indeed vital for industry to step in and develop products, software and services. One of the best results of government funding is, indeed, the creation of new industries. The Internet is perhaps one of the best examples, but the same is true in other areas, including many areas within computer science.
For example, take the development of the very important Ethernet. Xerox had created its research laboratory, the Palo Alto Research Center, or PARC, in the early 1970s. It started hiring some of the best and brightest from the DARPA research community. Bob Taylor ran one of the laboratories within PARC, so he knew most of the people. Bob Metcalfe had been at MIT and had worked on the ARPANET. He was one of the people Taylor recruited.
Bob Metcalfe came to visit me in Washington while I worked at DARPA, and I put him up for the night on the fold-out couch I had in my living room. I had brought home a paper describing the Alohanet in Hawaii, a radio-based network that used two channels (an uplink and a downlink) to connect terminals to a mainframe on the University of Hawaii campus. Alohanet was supported in part by DARPA, and I had chatted with Norm Abramson, one of the professors running the project, when he visited the DARPA office. I was intrigued by their bold approach of having all the terminals transmit on the same frequency, thereby running the risk of interfering with each other. Instead of carefully preventing collisions, they compensated by requiring successful transmissions to be acknowledged. If a terminal did not get an acknowledgment, it would send its message again. This scheme works nicely as long as there isn’t too much total traffic.
I showed this paper to Bob Metcalfe, who studied it before going to sleep. He took some issues with the math in the paper and he realized the same ideas could be applied to transmission in a cable. Thus Ethernet was born --a technology for connecting computers to each other in a local environment, usually an office or a laboratory--and I wound up with the unique privilege of being the host of its father.
Bob and his colleagues at PARC designed and implemented the Ethernet. After it was working well, Bob left PARC and with the cooperation of Xerox, Intel and Digital Equipment Corporation founded 3Com to make and sell Ethernet products. There were other approaches being developed for local area networks, but eventually Ethernet became dominant.
Ethernet and the creation of 3Com is just one of many, many examples of large companies coming into existence to cause the enormous expansion of the Internet. Cisco Systems, which started out as a router company, and Sun Microsystems, which developed some of the first Internet-enabled workstations, are two others.
Large-scale changes such as the Internet require many players and evolve over a fairly long period of time. I particularly like Bob Metcalfe’s formulation of the “Doriot Ecology” of innovation. This ecology is inhabited by research professors, graduating students, scaling entrepreneurs, venture capitalists, strategic partners, and early adopters, though I would have to add that enlightened government funding, exemplified by DARPA’s and later NSF’s role in spearheading network technology, also plays a strong role.
There’s no question that private industry has played a crucial role in expanding the Internet to its present point of serving billions of people. But there’s also no question that the government developed the basic technology and brought the Internet into existence.