How ARPANET Works

Before ARPANET, most computer systems consisted of a massive computer -- sometimes the size of an entire room -- with user terminals hardwired to it. A terminal was some form of user interface, often consisting of a keyboard or punch card reader. Multiple users could access the computer simultaneously, in a technique called timesharing. Other early networks required a direct connection between host computers, meaning that there was only one path for information to flow through. The direct connections limited the size of these computer networks, which became known as local area networks (LANs).

ARPA wanted to build a networked system that could stretch across the United States, linking governmental and scientific organizations in a way that had never been possible before. However, the first phase of ARPANET was much more modest: Four computer systems in different locations would link together using existing phone lines and four Interface Message Processors (IMPs).

ARPA chose the initial computer sites based on pre-existing research relationships with the United States government. Each site had its own team of engineers responsible for connecting the site computer to the ARPANET. The four host computers in the initial ARPANET structure included:

In August 1969, the UCLA team hooked up its host computer to an IMP, a Honeywell DDP 516 computer, making it the first of the four sites to connect into ARPANET. Within a few days, the two computers could exchange information. In October, Stanford's team added the second IMP and host to the system. At 10:30 p.m. on October 29, the Stanford and UCLA computers communicated with each other over a 50 kilobit per second (kbps) phone line.

On the first attempt, the system crashed before UCLA could send a complete command to the Stanford computer. Fortunately, everything worked on the second try. The other two host computers joined the network before the end of 1969. For the first time, scientists could harness the power of multiple computers in remote locations.

Most of us take the Internet for granted. We can log into our e-mail and browse the World Wide Web. But when the ARPANET project began, there were no procedures or systems in place to let different computers share information. Everything had to be invented from scratch.

One of the most important decisions the ARPANET team made was to create a standardized system of protocols that the host computers and IMPs would follow. A group called the Network Working Group formed to take on the job. At first, the creation process was chaotic and disjointed. The ARPANET team began to design and establish these protocols through its RFCs.

Early in the process, the team recognized the need for two specific tasks: Create a way for users to log in to the system remotely and make it possible to move files from one machine to another. Remote login later became known as Telnet, and moving files back and forth became part of the File Transfer Protocol (FTP). The team submitted these first two protocols to Larry Roberts, the head of the project. Roberts decided that the initial protocols weren't ambitious enough and told the team to include more functions and processes.

The team began to design the Network Control Program (NCP), a symmetric host-host protocol. In simpler terms, this was the procedure for allowing computers to communicate within the network as well as adding more hosts to the network to make it bigger. The NCP also controlled the path and flow of data over ARPANET. It established the practice of using numeric host addresses for network communication and was a forerunner to today's domain name servers (DNS).

ARPANET also took advantage of a revolutionary new way to send data: packet switching. In packet switching, host computers divide each computer file into smaller segments called packets. Once the packets are transferred, the pieces are reassembled into the original files.

Packet switching plays an enormously important role on the Internet today. Information can flow through the entire system much faster when computers break files into smaller pieces. That's because one huge file traveling through a pathway takes up a lot of bandwidth. Several smaller packets can travel through different paths within a network to the same destination. If part of a pathway is closed for some reason, the road-blocked packets can find another path while the rest of the information continues to load on the user's computer.

ARPANET allowed people to do things with computers that had never been done before or were only possible on a much smaller scale, including:

Shortly after e-mail debuted on ARPANET, a few team members began to play around with mailing-list software applications. Soon someone on ARPANET could send out an e-mail message to a group of people in one step. Mailing lists focused on specific topics began to appear. The first was SF-LOVERS, a list that connected the science fiction fans on the ARPANET team. But ARPA frowned on people using the network for unofficial purposes and demanded that list owners dissolve all unauthorized mailing lists. Users later convinced ARPA to allow these lists by pointing out that they helped test the network's mail capacity [source: American Heritage Magazine].

In 1973, Robert Kahn initiated an experiment with a technique he called internetting -- combining two or more separate networks into a larger network. He began to look into ways to integrate ARPANET with the Defense Advanced Research Projects Agency's (DARPA) Packet Radio Network, which was a network that used radio waves to send data from one computer to another.

In 1983, ARPANET officially switched from NCP to the TCP/IP suite of protocols. Just as ARPANET's architecture and protocols foreshadowed the Internet's structure, the applications users created to navigate and interact with the network paved the way for many of today's Internet features.

Between 1969 and 1977, ARPANET grew from a network of four computer sites to one with 111 computers belonging to universities, research facilities and the military. Using satellite links, ARPANET connected computer systems in the continental United States to computers in Hawaii and Europe. Even though the network had grown, few people actually had access to the system. In general, the public remained unaware of ARPANET's existence.

Other ARPANET networks began to go live, including USENET, Ethernet, CSNET and BITNET. The ARPANET Request for Comments 827 established an External Gateway Protocol that made it possible for separate networks to access each other, even though access to ARPANET was still restricted for official use. In 1983, the military section of ARPANET split off from the network; its only connection to the larger network was a few e-mail gateways. The military renamed its smaller network MILNET, which would later become part of the Department of Defense Data Network (DDN) [source: Living Internet].

In 1986, five supercomputer centers formed a network called NSFNET. Before long, NSFNET grew to include several universities in its network. Other networks began to consolidate into larger systems. People referred to this larger collection of networks and gateways as the Internet. While the era of the personal computer began in the late 1970s, the Internet still remained a resource for universities, corporations and the government.

ARPANET's infrastructure was beginning to show its age. The system's IMPs weren't as efficient or powerful as the computer nodes in other networks. Organizations on ARPANET began to transition to other networks, mainly NSFNET. In 1990, DARPA pulled the plug on the ARPANET project. The organization's goals had been met. The United States had a nationwide computer network that not only linked powerful resources together, but also could continue operating if a significant portion of the network stopped working. Even more impressive, this network now spanned the globe, connecting computers from one side of the world to the other.