Before there was the internet, there was the ARPANET. Ian Evenden speaks to some of those who were there, and discovers how they set about building something that had barely been imagined

© ISTOCK / GETTY IMAGES PLUS/DKOSIG

THE IDEA OF A NETWORK which spans the entire world is not a new one by any means. As far back as 1900, Nikola Tesla was dreaming of wireless power and message transmission, though sadly despite his claims, the chances are the wireless power part of the system would never have worked.

Thinkers and theorists kicked the idea around for much of the early 20th century, coming up with ways that information could be stored and accessed long before the technology to build them was actually available. MIT computer scientist and later ARPA office head, JCR Licklider, put forward the somewhat optimistic idea of an ‘intergalactic computer network’ in the early 1960s, but his work did a lot to ground the idea in reality instead of science-fiction. His ideas about an all-encompassing computer network contain almost everything the internet is today, including the cloud, simple user interfaces, AI, and ‘time sharing’—the idea of a central server accessed via several terminals at the same time.

Another concept is that of packet switching, breaking a stream of data down into parts, each with a header detailing where it’s going, that allows network hardware to make sure it ends up in the right place. This allows multiple computers to communicate on a single network, the data passing through routers that can decide the best route. This again came about in the 1960s, as part of research into fault-tolerant networking at the RAND Corporation, funded by the US Department of Defense, and through independent work by British scientist Donald Davies, who coined the term ‘packet switching’.

Only once all these concepts are in place do you have the origins of the ARPANET, the precursor to today’s internet. In 1966, at the United States Department of Defense’s Advanced Research Projects Agency, or ARPA—better known today as DARPA, the Defense Advanced Projects Agency—a project was initiated to allow access to remote computers. Much of the work was carried out at UCLA and Stanford Universities.

Today, we have a network of networks, known as internetworking, but it took until 1969 to connect the first two computers. Work continued on protocols, culminating in the Network Control Program, led by Steven Crocker. NCP was the predecessor to TCP/IP, which was proposed by Vint Cerf and Robert Kahn in the early 1970s. The document setting out its initial specification contains the first use of the word ’internet’. Other protocols developed by Crocker’s Network Working Group include TELNET, FTP, and SMTP. Finally, in 1983, ARPANET transitioned to using TCP/IP, and the modern internet age began, capped off by the invention of the World Wide Web by Sir Tim Berners-Lee at CERN in 1990.

However, in this feature, we’re going to concentrate on the early structure and hardware rather than the front-end of the internet. Building the ARPANET required many different technologies to mesh in ways that, up to this point, hadn’t been considered. When you’re striking out into unknown territory, failures are to be expected, but the story of the ARPANET and internet is one of gradual improvement and expansion over time, once the initial period of explosive innovation was out of the way. We may have Web 2.0 and Web3, but down below, the cables and routers remain much as they were.

Amusingly for anyone who’s ever used a computer, at least, the first message transmitted over the ARPANET in 1969 was truncated by a hardware failure. Only the ‘LO’ characters of LOGIN arrived at the target computer in Stanford, California, having been sent from UCLA, 350 miles away along the I5, with a full connection coming an hour later after some frantic debugging. Let’s hope we make it through the next six pages.

ARPA WAS NEARLY BEATEN TO IT by the British, who from 1968 had a local area network of their own, the NPL Data Communications Network operated in London’s National Physical Laboratory. It was based on designs by Donald Davies, the inventor of packet switching, and operated until 1986. Davies’ ideas included a national commercial data network and, in 1966, his team produced a design for the laboratory network, including the concept for an ‘interface computer’ we’d call a router.

The NPL ideas were presented at the Symposium on Operating Systems Principles, a conference held in 1967 in Gatlinburg, Tennessee. The plan for ARPANET was also presented there, by Larry Roberts of ARPA, and he took away with him many of Davies’ concepts, including the speed at which it should operate. ARPANET was originally designed to run at a transmission rate of 2.4kbit/s, but this was upgraded to 50kbit/s after NPL’s remarkable 786kbit/s line speed.

Independently, Paul Baran of the RAND Corporation had come up with a similar idea to packet-switching, published in his papers titled ‘On Survivable Communications’, but called the packets ‘message blocks’. However, his work lacked Davies’ breakthrough idea that computer network traffic comes in bursts, not a constant flow like a phone conversation. It was this, along with the ability of the routers (known as IMPs, or Interface Message Processors) to route traffic around failures as long as a path could be found, that fed into one of ARPANET’s disputed design goals. Baran’s work was meant to survive and adapt in the event of a nuclear attack. Was ARPANET?