Illuminating the net's Dark Ages

By Colin Barras
Technology reporter

Imagine a history of World War 2 that failed to cover the events in 1930s Germany. Conventional histories of the internet are that incomplete, according to a researcher.

It is difficult to conceive of a world without the internet - today more than a billion users worldwide are connected - but just 25 years ago global network connections were vanishingly rare.

During the 1980s, small research networks linking a few hundred universities were gradually replaced by a commercial network with 300,000 users.

But history has failed to document this transitional period in any detail. Dr Doug Gale, president of Information Technology Associates, in Montana, is devoting his spare time to filling in the gaps.


In the early 1980s, Dr Gale worked as a network administrator at the National Science Foundation (NSF), a US government agency that promotes research and education.

He says that at the time, there was just one research network, called Arpanet. It was run by the Defense Advance Research Projects Agency within the US Department of Defense.

Computer networking

But access to Arpanet was limited to universities with connections to the Department of Defense. As the advantages of computer networking became apparent, other universities began developing their own networks.

Ultimately, the NSF launched NSFNET, the immediate forerunner of the modern internet, to link all the networks together. But there are so few historical records from that time that Dr Gale describes the period as the internet's Dark Age.

"The Arpanet period is somewhat well documented because the corporation in charge - BBN - left a physical record," says Dr Gale. "Moving into the NSFNET era, it became an extraordinarily decentralised process. The record exists in people's basements, in closets."

Dr Gale has a simple explanation for the decentralised nature of the NSFNET era.

It was such an exciting project that lots of people just wanted to drop what they were doing and get involved
Doug Gale on the development of NFSNET

"It was such an exciting project that lots of people just wanted to drop what they were doing and get involved," he says.

There were collaborations between universities, governments, corporations and individuals in the private sector, Dr Gale explains. Although that led to rapid progress in the development of the internet, it has left a poor historical record.

"So much of what happened was done verbally and on the basis of individual trust," says Dr Gale. "If something needed to be done, there was never a formal contract; it was just done on the basis of a telephone call."

He says the more he explores the missing history, the more complicated the picture becomes.

"I thought I had some sense of what was going on. But it's a much richer history than I ever dreamed," he says. "A number of individuals were making important contributions in multiple areas."

Standard language

One of the most important of those individuals was Dr Dennis Jennings, now director of Computing Services at University College Dublin.

In the early 1980s, Dr Jennings made the key decisions at the NSF that would shape the modern internet. One of those decisions was to insist on a standard language, or set of protocols, to allow easy communication between networks.

Dr Jennings settled on the TCP/IP protocols, which had been developed by Robert Kahn and Vinton Cerf at Arpanet in the 1970s.

"I became known as 'Mr TCP/IP' because I was going around saying it has to be based around these protocols," Dr Jennings says.

"The decision to go with TCP was radical," he continues. "It was a great decision technically, but implementation of the software was poor and the initial performance was dreadful. But at the time we envisioned a few hundred users, not ten thousand."

Gradually, the software improved and more networks switched to TCP/IP.

But having a number of isolated networks speaking the same language wasn't enough. There needed to be a central backbone to connect the networks together. Moves began to form a network of networks, the first internet.

"That was the last piece of the jigsaw," says Dr Jennings.

The impetus for an internet came from the academic research community and a feeling that resources needed to be more evenly distributed, according to Dr Jennings.


"There was concern that some US researchers were not getting enough access to supercomputers," he says.

The NSF thought about a network to connect a handful of supercomputers, but Dr Jennings pressed for a more radical solution. In September 1985, he suggested creating a general purpose 'backbone' to connect any network user with any other user.

"The supercomputer centres thought it was an appalling idea," he says. "But then they caught on to the idea that if there was a backbone, everyone would be able to access their network."

The thought of all that extra traffic and revenue was enough to persuade the supercomputer centres to support Dr Jennings' calls for a general purpose network. The NSF launched NSFNET in July 1986.

That was a crucial turning point, says Dr Gale.

"There was an understanding that, although NFSNET was billed as a supercomputer network, that would soon become a less important component," he says.

Central backbone

"And within months, if not weeks, supercomputers did become just a small part of NSFNET. The rest was - well, you know the internet - everything else."

In Europe too, there were calls for a central backbone to connect existing networks. But the picture was more complex, says Dr Kees Neggers, managing director of Dutch research network SURFnet.

Public networks in Europe were running a set of protocols called X.25 rather than TCP/IP, he says.

Soon the networks could no longer handle the demand and consensus emerged among users to start a TCP/IP backbone in Europe.

In 1991, Dr Neggers chaired a meeting in Amsterdam which helped establish a TCP/IP network called Ebone, which would become the European equivalent of NSFNET.

In both Europe and the US, TCP/IP was viewed as an interim communication solution, to be replaced with superior protocols in time, says Dr Neggers. But by the mid 1990s, TCP/IP had become so ubiquitous that it was never replaced.

The internet had become commercial by this time, according to Dr Gale. The turning point came in 1987 when the NSFNET backbone was upgraded to 1.544 Megabits per second (Mbps).

Although slow by today's standards - broadband connections of 24Mbps are now available to UK users - it was fast enough to encourage commercial organisations to take a more active role in running the emerging internet. By 1990, the Internet boasted 300,000 users and numbers have been growing ever since.

"It didn't hit home at a gut level the changes it would have for culture and society. I saw it as a scientist; I didn't see it as a human," says Dr Gale.

Dr Gale has barely scratched the surface of internet history, but in the future he intends to extend the focus of his studies. Once he has a good record of internet history in the USA, he will look at the history in Europe and elsewhere in the world. He will also extend the record further into the past.

"When I started the history project the initial focus was that Dark Age in the 1980s and early '90s," he says.

"But the archive is not going to focus on a particular era. There's more to human history than the Dark Ages."
Story from BBC NEWS:

Published: 2007/08/23 07:08:11 GMT