Notes from Conversation with Professor David Farber:
A Pioneer’s Perspective on the Internet’s Beginnings and Growth
The Early Days of the Internet as the Experiment
In the early days of the Internet, it was literally an ‘experiment.’ Hardly anyone had computers, and there was no clear vision of where things were headed. For example, the DCS(Distributed Computing System) experiment was driven by the use of small computers, which were unreliable and had limited power. In a sense, there was no broader vision for society. Such experiments were purely motivated by computer scientists’ curiosity to see what would happen if connected computers—just give it a try and see it happen. Since a network is like a massive software, these experiments could be driven by the curiosity of software engineers.
There was the transition that the ‘experiment’ began to gradually take on a societal character, with the arrival of CSNET(Computer Science Network) which went from computer researchers. These researchers can be said the first real users of the network, and they tended to be more tolerant in general. CSNET wasn’t governed by the computer science community because computer scientists saw the network as just a tool—they were satisfied as long as it worked. Instead, a small group on the advisory board effectively and strongly ran things. The NSF(National Science Foundation) director and advisory board were thought to be heading in the right direction. And, there was a simple law, I’d call this the first and simple ‘law’ of the Internet, in a sense that ‘if you violated it, you are kicked off’ among these researchers.
In this way, I started network research as an experiment driven by the pure motivation of a researcher. As regional networks grew, stakeholders beyond just computer scientists began to emerge, and it was at that point I started thinking about adapting the network to society. And, when Bob Kahn and I proposed the concept of a gigabyte network, I sensed the future of the network. In this way, the network has evolved.
The Roles of Regional Networks
When we started building traffic in the U.S., university computer science departments were not prepared enough to connect the nation. The U.S. is a very big country, and telephone calls across the U.S. were extremely expensive. Therefore, we encouraged the regional consortium, cooperation from regional industry, government aid, and support universities and connect regions to regions. We tried to regionalize the problem because each region had a different problem. Then, work it well. Throughout this period, there were no rules except for the restrictions on the use of government funds.
Changes That Came with the Commercialization of the Internet
The commercialization of the Internet wasn’t as organized as it should have been because it happened almost accidentally in many ways. Many mistakes were made during that time. I wish it had worked better. The U.S. government, including agencies like the FCC(Federal Communications Commission), should have been interested in the Internet, but they didn’t—they saw it as just a toy. The telephone industry also failed to recognize its potential.
Initially, as I said, the network was restricted in the academia. However, some commercial companies like IBM lab, which wants to supply computing equipment for universities, requested to get on the network. So, we essentially said, ‘Sure, of course, but you have to pay the annual fee that the university pays several times.’ In this way, companies with a clear commercial purpose were able to rationalize the costs of connecting. When the Internet expanded outside of the United States, the situation was roughly the same. The cost of connecting the Internet to countries like Japan and South Korea was ridiculously high.
The Expansion of the Network
Additionally, when connecting to other countries, we had a simple rule: we only talked to one group from one country. For instance, in Korea, two people came to us, and they claimed they represented Korea, respectively. If a representative couldn’t be chosen within a country, they couldn’t connect to us. Thus, this simple rule would be the initial rule for national control.
When the network was connected to Japan and South Korea, it was still linked with computer scientists, so I didn’t think the nature of CSNET had changed significantly. We were all out of the same box. However, the endless problem was the language. For example, discussions about how to structure the codes should be to facilitate communication between Japanese and English took a lot of time. However, people began to realize that fundamentally the Internet could quickly spread knowledge. So, in the U.S., it started to become a big issue, particularly from the perspective of encryption.
What Made the Internet So Popular
As I said, in the early days of the network, it was for academic researchers, and they were using mainframes like PDP-11. It was not something for the average person. Therefore, there was naturally limited growth of the network. And then, IBM announced the Personal Computer, which was the first real computer. They did a very good job, but the problem was the network. The code written for PDP-11 in MIT to run the network pulled it over to run IBM PC. Once that happened, everybody was getting copied. And again, the market developed very fast.
The mechanisms established by the IETF(Internet Engineering Task Force) were also remarkably good. Efficient decision-making through rough consensus allowed for quick action. While these may not necessarily be considered the ‘laws’ of the Internet, they were very effective in practice. In contrast, the ISO(International Organization for Standardization) created a beautiful standard, but it didn’t result in something that worked in practice.
As for whether the web was a game changer, it can be considered more of a game enhancer.
Fundamentally, if we consider that each computer is connected through the network, we can think in terms of layers: the hardware, the network, and the application layer. So, many of today’s advancements, such as AI, are recognized as issues at the application layer. However, when considering security, we must rethink it starting from the hardware layer.
Why Security Is Crucial and Why It Was Challenging
One of the biggest mistakes was that we paid very little attention to security. In the early ages of the network community, it was hard to make it work because we knew each other and nobody was going to crash the network, so security never should’ve happened. We didn’t care about security. NSFNET basically we had a minimum of security mechanisms.
However, when I ran the first email system at the University of Delaware, I established my own set of rules for the staff: ‘If you looked at somebody’s mail, you’re gonna get fired.’ This rule was in place because there was no encryption, and the computer was too slow.
From the perspective of the 1970s, I could say that today’s network has proven to be effective, especially in the sense that it was ready in time for the COVID-19 pandemic. In particular, platforms like Zoom (similar to See You See Me) where low latency, have been incredibly useful.
However, if I were much younger, one of the things I would want to focus on—despite having organized some large projects—would be redesigning the network from a security perspective.
First, the hardware layer is buggy, but now we know how to design secure machines. Second, on the software layer, nobody truly knows how it works with OS. They have become just too complicated and too large, although when I was building OS, I could tell you how it is organized in detail, so there are holes. It is almost a miracle that we haven’t had more problems. Ultimately, as we depend on more and more computers, we become more vulnerable and it can be pretty serious.
Notes from Conversation with Professor David Farber
Author: Hitomi Sano
Location: Roppongi Hills Club, Tokyo, Japan
Meeting Date: February 4, 2025