The Genesis of the New Internet
by Mark Bayliss
CEO, Visual Link Internet
and
Christopher Harz
VP, IPv6 Summit

As we start presenting the New Internet, IPv6, to a wider audience, we've started getting more general questions. When we start comparing IPv6 to IPv4, we hear the question, "But what is IPv4? Where did all this start?" This is one of a series of background articles in answer to that question.

To understand the present Internet, and some of its strengths and limitations, it is useful to remember the time when it was created, in the 1960s, at the height of the Cold War, when there were two Superpowers in the world, both with enough nuclear might to kill each other several times over. Vast armadas of tanks, missiles and other military equipment were lined up facing each other across the East German border, each commanded by generals worried that the other side would "get the jump" on him – that his side would be the victim of a surprise attack. This was before the advent of personal computers, distributed processing, or widespread use of satellite links.

Communication was by relatively few lines that passed through even fewer nodes – which were vulnerable to attack. Computing was done via centralized hosts, each of which ran on proprietary systems – in other words, if one headquarters computer center were to be knocked out by a missile, another could not step in to take its place, and every headquarters downstream of that center would be left in the dark, information-wise, wondering whether a sky full of missiles was headed its way, and whether it should launch its forces first.

The nation had just received a rude surprise, in 1957, with the launch of Sputnik by the Soviet Union. The perception that we were falling behind the Soviet Union in technology led to the formation of ARPA, the Advanced Research Projects Agency, by the Pentagon – a centralized, highly streamlined group of smart project managers that could start projects and fund them within weeks or months, in contrast to other Pentagon research centers, which could often take more than five years to get a project off the ground. One of the first projects ARPA tackled was the "single point failure" vulnerability of military command networks – the troubling fact that eliminating one point in a structure could cause the whole system to fail. Dr. Larry Roberts, one of the "Fathers of the Internet," became the project manager of this initiative at ARPA, funding and directing the project that was eventually to lead to the "Arpanet."

The basic concept of the Arpanet had been started by Paul Baran at the RAND Corporation, the military think tank founded by 5-star Air Force General Hap Arnold. Baran, postulated breaking information into small blocks of data using "digital" and "distributed" methods – novel concepts at the time. Every item of information, such as a written or voice message, would be broken down into small, similar blocks. Each block of data would be in digital (i.e., ones and zeros, as opposed to analog) form, and have a forwarding and a return address, much like the envelope of a letter, and a payload of information, much like the contents of the letter itself. Each such block of data could take a different path from sender to receiver – the total number of paths would be distributed throughout the entire network.

This is in contrast to a normal telephone message via a switched circuit, which ties up a dedicated path from sender to receiver for the duration of the message. The dedicated path method is relatively inefficient – for instance, for a voice message, people don't speak very quickly, and there are many pauses in the speech; by contrast, breaking up the speech into small blocks and speeding up the throughput allows many voice transmissions to take place at the same time. In addition to being more efficient, the distributed transmission is more likely to get there. We can think of a classic computer system like a star (figure 1) – if you knock out the center of the star, the entire system is dead. By contrast, Baran's concept is like a fishnet, with each knot being a node in the network (figure 2); if one node is down, you simply route around it.

Figure 1: Star System		Figure 2: Mesh System

Under this network concept, each node would receive a block of information and then quickly forward it towards the receiver via whatever path lay open – Baran called this "hot-potato routing." A researcher named Davies, who had worked independently on such networks in Great Britain (as had Kleinrock and others at MIT) baptized this method with the more elegant name of "packet switching."

With the concept formulated, it now remained to actually get computers to send such packets. This took place at the University of California at Los Angeles (UCLA), under the guidance of Dr. Leonard Kleinrock, whose team sent the first packets from UCLA to Stanford Research Institute in 1969. The first sketch of the system is presented as figure 3.

Figure 3: The First Internet Network

In addition to the basic format of the packet of information, a standardized means of routing the packets was also necessary, and this was provided by two further Fathers of the Internet, Drs. Vincent Cerf and Robert Kahn, who invented the Transmission Control Protocol/Internet Protocol (TCP/IP). The system of computers that sent these first messages back and forth and would eventually be called the Internet was first called the "Arpanet," and, by 1971, grew to the configuration shown in figure 4. It then expanded slowly in America, mostly on the two coasts (figure 5), and then started expanding worldwide in the mid 1980s (figure 6). The genesis of the Internet is still evident in today's traffic patterns – much of the global Internet traffic still passes through northern Virginia, within a few miles of the headquarters of ARPA (now called DARPA, the Defense Advanced Research Projects Agency), where it all began (figure 7).

After the concept had been physically demonstrated, one would suppose that other university and military computer centers would have been clamoring to get on board, but this was not so. "There was a lot of resistance to joining the Arpanet, at first," said Dr. Roberts. "Managers of data centers saw no need to share their data, and also did not want to reconfigure their proprietary architectures." It took some arm twisting to get these centers to fall in line. "I simply told them that if they did not join the network that they would receive no further ARPA funding," Dr. Roberts noted. "That was usually enough to convert them over and make believers out of them."

Figure 4: The Arpanet in 1971

Figure 5: The Arpanet in 1980

Figure 6: The "USENET" in 1986

Figure 7: Today's main Internet trunks. Note the peak (coded red) traffic in Virginia in the US.

The TCP/IP standard was adopted by the entire nascent Internet on Jan. 1, 1983, a "flag day" when all the computers switched over. The standard had already been adopted by the U.S. Department of Defense three years earlier. This relatively sudden and community-wide transition to a new common version of the Internet stands in contrast to the situation today. There is no worldwide authority that has the financial or other power to cause a switch from the existing version 4 protocol to the new version 6 protocol – hence both will co-exist (with concomitant compatibility problems) for years to come. Individual countries such as South Korea and Japan appear to be the exception, with the strong national focus and financial incentives to make the transition happen in a planned and speedy manner, reminiscent of Dr. Robert and his remarkable leadership in the early days of the Internet.

In summary, then, one can see many of the strengths and weaknesses of today's Internet in the kernel of the early Arpanet. It was originally formed as a communications system between relatively few university and military centers – there was no thought of wide public usage, with concomitant problems such as spam and viruses. The address length (32 bits) of each Internet packet was thought to be enough for all possible future users – in fact, with about four billion different addresses theoretically possible, it was thought to be overkill. It was not possible to imagine that a time would come when Internet addresses would be needed not only by most of the people on Earth, but also by billions of devices such as cell phones; the hoarding of huge blocks of Internet addresses by certain institutions, which would reduce the number of actual addresses available globally, was also not foreseen.

The early Internet was intended for short messages – huge blocks of content were undreamed of. Internet nodes were meant to be located at fixed sites such as military bases – again, having millions of mobile Internet users was far beyond the scope of what was considered possible at the time. The network from the beginning had a tradition of supporting R&D for new uses – for instance, an early node, the University of Utah, featured Dr. Sutherland working on 3D graphics transmitted over the Net. But, again, the creative recent social uses of the Internet, such as the vast virtual worlds of today's online gaming environments (e.g., World of Warcraft and Second Life), were undreamed of.

In future articles in this series we'll explore how the New Internet was designed to overcome the shortcomings of the existing Internet's 30-year old architecture, and what new "killer apps" lie right around the corner.