From the beginning, the thing that set the early Internet apart from other communication systems was its use of packet switching. Information was delivered in packets rather than as a single continuous transmission. If a glitch occurred during transmission, it was only necessary to resend the packets affected by the glitch.

Leonard Kleinrock developed the packet concept at MIT in the late 1950s and published a theory paper in 1961. Kleinrock later became a key figure in the construction of the early Internet. About the same time, Paul Baran of Rand Corporation developed a model of a distributed mesh topology network. He also developed the concept of multiplexing, in which packet transmissions from multiple sources may be traversing the network at the same time.

Baran was working on a U.S. Department of Defense project to build a communications network that could survive a variety of disruptions, including nuclear war. In thinking about the brain's neural network of redundant connections, he came up with the distributed mesh design shown on the right in the following illustration.

The centralized model was common with mainframe systems of the time. Dumb terminals connected to central computers that controlled everything. The decentralized model was basically a group of mainframes that could exchange information. The distributed mesh model supports any-to-any connectivity, meaning that any node can directly connect with any other node. This requires a universal logical addressing scheme and a mechanism that allows packets to flow from source to destination essentially unaltered.

The network designers realized that there were a variety of different hosts to connect, so a separate message-handling device was created to handle the communication process in a standard way and provide a custom interface to the local time-sharing host system. The device was called an interface message processor or IMP, which was the forerunner of today's router. The first IMP went online in 1969, providing communication services between computers at UCLA and Stanford. At this point, the early Internet was really just a remote computing network, not an internetwork. But it would soon grow and be referred to as the ARPANET.

The early designers of the Internet were particularly interested in the concept of "open architecture networking." People working on the project wanted to connect with many different types of computer systems over many different types of transmission schemes, including low-speed, high-speed, and wireless connections. The IMP separated communication services from the actual computer. This allowed any type of device to connect to the network. The result of this open approach can be seen in the near universal connectivity provided by the Internet.

Bob Kahn of DARPA and Vinton Cerf at Stanford University began developing TCP (Transmission Control Protocol) in the early 1970s. The original protocol was called TCP (IP wasn't defined until later) and it provided a range of reliable connection-oriented services (flow control, acknowledgment, retransmission, and so on). But the designers soon realized that TCP's reliability features added overhead that disrupted the ability to deliver live voice across the network. So, in 1978, TCP was reorganized into TCP and IP, with TCP providing reliability, and IP handling basic networking functions such as addressing, routing, and packet forwarding. With this, applications that didn't need TCP's reliability functions could bypass it and go directly to IP through a new protocol called UDP (User Datagram Protocol). UDP is a scaled-down version of TCP. It provides port connections for applications but foregoes all the extra services in the interest of speed.

The official "birth" of the Internet and the transition from the ARPANET occurred on January 1, 1983. This was the day that all connected networks were required to run the new TCP/IP protocol suite. While the Internet has a history stretching back over 20 years, from that day on, it was called the "Internet."

ARPANET grew during the 1970s, partly because its open architecture made it easy to connect just about any system to the network. In fact, the more computers that were added, the better-in the same way that fax machines are more beneficial if everybody owns one. E-mail was probably one of the most significant contributors to Internet growth. Eventually, networks (and not just large computers) were attached to the network. Robert Metcalfe had created Ethernet, which immediately became a popular LAN technology. Also, the TCP/IP protocols were adopted by the UNIX community, which led to a rapid explosion in the number of users that could connect to the Internet.

In the early 1980s, the NSF (National Science Foundation) realized the benefits the ARPANET had on research and decided to build a successor to ARPANET that would be open to university research groups. Its project resulted in the creation of a high-speed backbone called the NSFNET, which changed the topology of the Internet from a distributed mesh to a hierarchical scheme in which regional networks connected to the backbone and local networks connected to the regional networks.

This model proved very successful and it is still with us today, except that the hierarchical model has reverted to a fully meshed model as new backbones were built and interconnected with one another. In addition, regional backbone operators built cross-links directly between their networks rather than going through the backbone. By 1995, the NSF had defunded NSFNET and implemented full commercialization of the Internet. See "Internet Architecture and Backbone" and "ISPs (Internet Service Providers)" for more information.

In 1996, the Next Generation Internet (NGI) Initiative was announced to develop advanced networking technologies and applications on testbed networks that were 100 to 1,000 times faster than the networks of the day. This research is still continuing, and you can learn more at the NGI Web site at http://www.ngi.gov.

Internet2 is the latest testbed for the Internet. It is a collaborative project sponsored by UCAID (University Corporation for Advanced Internet Development), a consortium of over 180 U.S. universities that are developing advanced Internet technologies and applications to support research and higher education. See "Internet2."

The core of the Internet is now based on optical networking technologies and light circuits or lambdas. A lambda is a single wavelength of light that can carry huge amounts of data. Multiple lambdas may exist in a single fiber strand thanks to DWDM (dense wave division multiplexing). These circuits provide high-speed point-to-point links that cross entire continents, thus eliminating router hops and avoiding congestion. High-speed optical access networks are also being developed in metropolitan areas, displacing the need to lease voice lines in order to carry data. See "WDM (Wavelength Division Multiplexing), "MAN (Metropolitan Area Network)," "Network Core Technologies," "Network Access Services," "NPN (New Public Network)," and "Optical Networks."