Jul 12, 2021
The Internet is not a simple story to tell. In fact, every sentence here is worthy of an episode if not a few.
Many would claim the Internet began back in 1969 when the first node of the ARPAnet went online. That was the year we got the first color pictures of earthen from Apollo 10 and the year Nixon announced the US was leaving Vietnam. It was also the year of Stonewall, the moon landing, the Manson murders, and Woodstock. A lot was about to change.
But maybe the story of the Internet starts before that, when the basic research to network computers began as a means of networking nuclear missile sites with fault-tolerant connections in the event of, well, nuclear war. Or the Internet began when a T3 backbone was built to host all the datas. Or the Internet began with the telegraph, when the first data was sent over electronic current. Or maybe the Internet began when the Chinese used fires to send messages across the Great Wall of China. Or maybe the Internet began when drums sent messages over long distances in ancient Africa, like early forms of packets flowing over Wi-Fi-esque sound waves.
We need to make complex stories simpler in order to teach them, so if the first node of the ARPAnet in 1969 is where this journey should end, feel free to stop here. To dig in a little deeper, though, that ARPAnet was just one of many networks that would merge into an interconnected network of networks. We had dialup providers like CompuServe, America Online, and even The WELL. We had regional timesharing networks like the DTSS out of Dartmouth University and PLATO out of the University of Illinois, Champaign-Urbana. We had corporate time sharing networks and systems. Each competed or coexisted or took time from others or pushed more people to others through their evolutions. Many used their own custom protocols for connectivity. But most were walled gardens, unable to communicate with the others.
So if the story is more complicated than that the ARPAnet was the ancestor to the Internet, why is that the story we hear? Let’s start that journey with a memo that we did an episode on called “Memorandum For Members and Affiliates of the Intergalactic Computer Network” sent by JCR Licklider in 1963 and can be considered the allspark that lit the bonfire called The ARPANet. Which isn’t exactly the Internet but isn’t not. In that memo, Lick proposed a network of computers available to research scientists of the early 60s. Scientists from computing centers that would evolve into supercomputing centers and then a network open to the world, even our phones, televisions, and watches.
It took a few years, but eventually ARPA brought in Larry Roberts, and by late 1968 ARPA awarded an RFQ to build a network to a company called Bolt Beranek and Newman (BBN) who would build Interface Message Processors, or IMPs. The IMPS were computers that connected a number of sites and routed traffic. The first IMP, which might be thought of more as a network interface card today, went online at UCLA in 1969 with additional sites coming on frequently over the next few years. That system would become ARPANET.
The first node of ARPAnet went online at the University of California, Los Angeles (UCLA for short). It grew as leased lines and more IMPs became more available. As they grew, the early computer scientists realized that each site had different computers running various and random stacks of applications and different operating systems. So we needed to standardize certain aspects connectivity between different computers.
Given that UCLA was the first site to come online, Steve Crocker from there began organizing notes about protocols and how systems connected with one another in what they called RFCs, or Request for Comments. That series of notes was then managed by a team that included Elizabeth (Jake) Feinler from Stanford once Doug Engelbart’s project on the “Augmentation of Human Intellect” at Stanford Research Institute (SRI) became the second node to go online. SRI developed a Network Information Center, where Feinler maintained a list of host names (which evolved into the hosts file) and a list of address mappings which would later evolve into the functions of Internic which would be turned over to the US Department of Commerce when the number of devices connected to the Internet exploded. Feinler and Jon Postel from UCLA would maintain those though, until his death 28 years later and those RFCs include everything from opening terminal connections into machines to file sharing to addressing and now any place where the networking needs to become a standard.
The development of many of those early protocols that made computers useful over a network were also being funded by ARPA. They funded a number of projects to build tools that enabled the sharing of data, like file sharing and some advancements were loosely connected by people just doing things to make them useful and so by 1971 we also had email. But all those protocols needed to flow over a common form of connectivity that was scalable. Leonard Kleinrock, Paul Baran, and Donald Davies were independently investigating packet switching and Roberts brought Kleinrock into the project as he was at UCLA. Bob Kahn entered the picture in 1972. He would team up with Vint Cerf from Stanford who came up with encapsulation and so they would define the protocol that underlies the Internet, TCP/IP. By 1974 Vint Cerf and Bob Kahn wrote RFC 675 where they coined the term internet as shorthand for internetwork. The number of RFCs was exploding as was the number of nodes. The University of California Santa Barbara then the University of Utah to connect Ivan Sutherland’s work. The network was national when BBN connected to it in 1970. Now there were 13 IMPs and by 1971, 18, then 29 in 72 and 40 in 73. Once the need arose, Kleinrock would go on to work with Farouk Kamoun to develop the hierarchical routing theories in the late 70s.
By 1976, ARPA became DARPA. The network grew to 213 hosts in 1981 and by 1982, TCP/IP became the standard for the US DOD and in 1983, ARPANET moved fully over to TCP/IP. And so TCP/IP, or Transport Control Protocol/Internet Protocol is the most dominant networking protocol on the planet. It was written to help improve performance on the ARPAnet with the ingenious idea to encapsulate traffic. But in the 80s, it was just for researchers still. That is, until NSFNet was launched by the National Science Foundation in 1986.
And it was international, with the University College of London connecting in 1971, which would go on to inspire a British research network called JANET that built their own set of protocols called the Colored Book protocols. And the Norwegian Seismic Array connected over satellite in 1973. So networks were forming all over the place, often just time sharing networks where people dialed into a single computer.
Another networking project going on at the time that was also getting funding from ARPA as well as the Air Force was PLATO. Out of the University of Illinois, was meant for teaching and began on a mainframe in 1960. But by the time ARPAnet was growing PLATO was on version IV and running on a CDC Cyber. The time sharing system hosted a number of courses, as they referred to programs. These included actual courseware, games, convent with audio and video, message boards, instant messaging, custom touch screen plasma displays, and the ability to dial into the system over lines, making the system another early network. In fact, there were multiple CDC Cybers that could communicate with one another. And many on ARPAnet also used PLATO, cross pollinating non-defense backed academia with a number of academic institutions.
The defense backing couldn’t last forever. The Mansfield Amendment in 1973 banned general research by defense agencies. This meant that ARPA funding started to dry up and the scientists working on those projects needed a new place to fund their playtime. Bob Taylor split to go work at Xerox, where he was able to pick the best of the scientists he’d helped fund at ARPA. He helped bring in people from Stanford Research Institute, where they had been working on the oNLineSystem, or NLS and people like Bob Metcalfe who brought us Ethernet and better collusion detection. Metcalfe would go on to found 3Com a great switch and network interface company during the rise of the Internet.
But there were plenty of people who could see the productivity gains from ARPAnet and didn’t want it to disappear. And the National Science Foundation (NSF) was flush with cash. And the ARPA crew was increasingly aware of non-defense oriented use of the system. So the NSF started up a little project called CSNET in 1981 so the growing number of supercomputers could be shared between all the research universities. It was free for universities that could get connected and from 1985 to 1993 NSFNET, surged from 2,000 users to 2,000,000 users. Paul Mockapetris made the Internet easier than when it was an academic-only network by developing the Domain Name System, or DNS, in 1983. That’s how we can call up remote computers by names rather than IP addresses. And of course DNS was yet another of the protocols in Postel at UCLAs list of protocol standards, which by 1986 after the selection of TCP/IP for NSFnet, would become the standardization body known as the IETF, or Internet Engineering Task Force for short. Maintaining a set of protocols that all vendors needed to work with was one of the best growth hacks ever. No vendor could have kept up with demand with a 1,000x growth in such a small number of years.
NSFNet started with six nodes in 1985, connected by LSI-11 Fuzzball routers and quickly outgrew that backbone. They put it out to bid and Merit Network won out in a partnership between MCI, the State of Michigan, and IBM. Merit had begun before the first ARPAnet connections went online as a collaborative effort by Michigan State University, Wayne State University, and the University of Michigan. They’d been connecting their own machines since 1971 and had implemented TCP/IP and bridged to ARPANET. The money was getting bigger, they got $39 million from NSF to build what would emerge as the commercial Internet.
They launched in 1987 with 13 sites over 14 lines. By 1988 they’d gone nationwide going from a 56k backbone to a T1 and then 14 T1s. But the growth was too fast for even that. They re-engineered and by 1990 planned to add T3 lines running in parallel with the T1s for a time. By 1991 there were 16 backbones with traffic and users growing by an astounding 20% per month.
Vint Cerf ended up at MCI where he helped lobby for the privatization of the internet and helped found the Internet Society in 1988. The lobby worked and led to the the Scientific and Advanced-Technology Act in 1992. Before that, use of NSFNET was supposed to be for research and now it could expand to non-research and education uses. This allowed NSF to bring on even more nodes. And so by 1993 it was clear that this was growing beyond what a governmental institution whose charge was science could justify as “research” for any longer.
By 1994, Vent Cerf was designing the architecture and building the teams that would build the commercial internet backbone at MCI. And so NSFNET began the process of unloading the backbone and helped the world develop the commercial Internet by sprinkling a little money and know-how throughout the telecommunications industry, which was about to explode. NSFNET went offline in 1995 but by then there were networks in England, South Korea, Japan, Africa, and CERN was connected to NSFNET over TCP/IP. And Cisco was selling routers that would fuel an explosion internationally. There was a war of standards and yet over time we settled on TCP/IP as THE standard.
And those were just some of the nets. The Internet is really not just NSFNET or ARPANET but a combination of a lot of nets. At the time there were a lot of time sharing computers that people could dial into and following the release of the Altair, there was a rapidly growing personal computer market with modems becoming more and more approachable towards the end of the 1970s. You see, we talked about these larger networks but not hardware.
The first modulator demodulator, or modem, was the Bell 101 dataset, which had been invented all the way back in 1958, loosely based on a previous model developed to manage SAGE computers. But the transfer rate, or baud, had stopped being improved upon at 300 for almost 20 years and not much had changed. That is, until Hayes Hayes Microcomputer Products released a modem designed to run on the Altair 8800 S-100 bus in 1978. Personal computers could talk to one another.
And one of those Altair owners was Ward Christensen met Randy Suess at the Chicago Area Computer Hobbyists’ Exchange and the two of them had this weird idea. Have a computer host a bulletin board on one of their computers. People could dial into it and discuss their Altair computers when it snowed too much to meet in person for their club. They started writing a little code and before you know it we had a tool they called Computerized Bulletin Board System software, or CBBS. The software and more importantly, the idea of a BBS spread like wildfire right along with the Atari, TRS-80, Commodores and Apple computers that were igniting the personal computing revolution.
The number of nodes grew and as people started playing games, the speed of those modems jumped up with the v.32 standard hitting 9600 baud in 84, and over 25k in the early 90s. By the early 1980s, we got Fidonet, which was a network of Bulletin Board Systems and by the early 90s we had 25,000 BBS’s. And other nets had been on the rise. And these were commercial ventures.
The largest of those dial-up providers was America Online, or AOL. AOL began in 1985 and like most of the other dial-up providers of the day were there to connect people to a computer they hosted, like a timesharing system, and give access to fun things. Games, news, stocks, movie reviews, chatting with your friends, etc. There was also CompuServe, The Well, PSINet, Netcom, Usenet, Alternate, and many others. Some started to communicate with one another with the rise of the Metropolitan Area Exchanges who got an NSF grant to establish switched ethernet exchanges and the Commercial Internet Exchange in 1991, established by PSINet, UUNet, and CERFnet out of California.
Those slowly moved over to the Internet and even AOL got connected to the Internet in 1989 and thus the dial-up providers went from effectively being timesharing systems to Internet Service Providers as more and more people expanded their horizons away from the walled garden of the time sharing world and towards the Internet. The number of BBS systems started to wind down. All these IP addresses couldn’t be managed easily and so IANA evolved out of being managed by contracts from research universities to DARPA and then to IANA as a part of ICANN and eventually the development of Regional Internet Registries so AFRINIC could serve Africa, ARIN could serve Antarctica, Canada, the Caribbean, and the US, APNIC could serve South, East, and Southeast Asia as well as Oceania LACNIC could serve Latin America and RIPE NCC could serve Europe, Central Asia, and West Asia. By the 90s the Cold War was winding down (temporarily at least) so they even added Russia to RIPE NCC.
And so using tools like WinSOCK any old person could get on the Internet by dialing up. Modems for dial-ups transitioned to DSL and cable modems. We got the emergence of fiber with regional centers and even national FiOS connections. And because of all the hard work of all of these people and the money dumped into it by the various governments and research agencies, life is pretty darn good.
When we think of the Internet today we think of this interconnected web of endpoints and content that is all available. Much of that was made possible by the development of the World Wide Web by Tim Berners-Lee in in 1991 at CERN, and Mosaic came out of the National Center for Supercomputing applications, or NCSA at the University of Illinois, quickly becoming the browser everyone wanted to use until Mark Andreeson left to form Netscape. Netscape’s IPO is probably one of the most pivotal moments where investors from around the world realized that all of this research and tech was built on standards and while there were some patents, the standards were freely useable by anyone.
Those standards let to an explosion of companies like Yahoo! from a couple of Stanford grad students and Amazon, started by a young hedge fund Vice President named Jeff Bezos who noticed all the money pouring into these companies and went off to do his own thing in 1994. The companies that arose to create and commercialize content and ideas to bring every industry online was ferocious.
And there were the researchers still writing the standards and even commercial interests helping with that. And there were open source contributors who helped make some of those standards easier to implement by regular old humans. And tools for those who build tools. And from there the Internet became what we think of today. Quicker and quicker connections and more and more productivity gains, a better quality of life, better telemetry into all aspects of our lives and with the miniaturization of devices to support wearables that even extends to our bodies. Yet still sitting on the same fundamental building blocks as before. The IANA functions to manage IP addressing has moved to the private sector as have many an onramp to the Internet. Especially as internet access has become more ubiquitous and we are entering into the era of 5g connectivity.
And it continues to evolve as we pivot due to new needs and threats a globally connected world represent. IPv6, various secure DNS options, options for spam and phishing, and dealing with the equality gaps surfaced by our new online world. We have disinformation so sometimes we might wonder what’s real and what isn’t. After all, any old person can create a web site that looks legit and put whatever they want on it. Who’s to say what reality is other than what we want it to be. This was pretty much what Morpheus was offering with his choices of pills in the Matrix. But underneath it all, there’s history. And it’s a history as complicated as unraveling the meaning of an increasingly digital world. And it is wonderful and frightening and lovely and dangerous and true and false and destroying the world and saving the world all at the same time.
This episode is pretty simplistic and many of the aspects we cover have entire episodes of the podcast dedicated to them. From the history of Amazon to Bob Taylor to AOL to the IETF to DNS and even Network Time Protocol. It’s a story that leaves people out necessarily; otherwise scope creep would go all the way back to to include Volta and the constant electrical current humanity received with the battery. But hey, we also have an episode on that! And many an advance has plenty of books and scholarly works dedicated to it - all the way back to the first known computer (in the form of clockwork), the Antikythera Device out of Ancient Greece. Heck even Louis Gerschner deserves a mention for selling IBM’s stake in all this to focus on things that kept the company going, not moonshots.
But I’d like to dedicate this episode to everyone not mentioned due to trying to tell a story of emergent networks. Just because they were growing fast and our modern infrastructure was becoming more and more deterministic doesn’t mean that whether it was writing a text editor or helping fund or pushing paper or writing specs or selling network services or getting zapped while trying to figure out how to move current that there aren’t so, so, so many people that are a part of this story. Each with their own story to be told. As we round the corner into the third season of the podcast we’ll start having more guests. If you have a story and would like to join us use the email button on thehistoryofcomputing.net to drop us a line. We’d love to chat!