A Compressed History of The Internet

It is true that the Internet is mysterious, but that will soon change (or so I hope). The key to unraveling the mystery lies in learning the history. But, before we start, let's peek at the state of affairs during the "birth" of the Internet in 1960.

In the 1960's, Ma Bell (Bell Telephone) had a monopoly on telephone lines and usage. Also, residential phone usage was subsidized by the business phone system, which meant higher costs for businesses using phone systems or phone lines.

The integrated circuit was yet to be invented, so the new-fangled computer systems of that day were manufactured using transistors and vacuum tubes by big companies like International Business Machines, Inc. and Digital Equipment Corporation. Since the computer was relatively new and few had been sold, most companies did not exploit the full power of the computer. As for most people, they scarcely knew what computers did. Also, since the computer had not matured, most companies could not imagine the power of linking computers together over large distances. Thus, any networking of computers and peripherals was done using proprietary techniques for encoding and transmission, much like the proprietary nature of Commodore's serial bus.

A little closer to home, Jack Tramiel of Commodore was expanding his typewriter business to include adding machines, which were electro-mechanical. The Commodore PET computer was not even an idea in Jack's head, since the necessary components had not been invented yet. Keep that in mind as we travel back to...

The 1960's
The United States and U.S.S.R are in the midst of the "Cold War", and each side is planning for a dreaded nuclear war. The U.S. government worries about how its offices and bases will stay connected after an attack. To search for answers, the government poses the question to America's foremost think-tank, RAND Corporation.

In 1962, Paul Baran of RAND Corporation describes a method of increasing reliability in telephone networks in the event of nuclear war. Out of his research comes the idea of breaking information into "packets" that can each be addressed to its destination. The best analogy would be letters at a U. S. Post Office. Each packet (letter) would contain a piece of the message, so the chances of parts of the message arriving at the destination would be increased. Of course, the receiver would have to request retransmission of any missing parts, but something was better than nothing. As the years progress, the construct of packetizing information results in the implementation of packet switching, a process where many people share a single data path. The packet switching unit splits the user's message into packets and addresses each packet. The packets are sent with everyone else's, and the unit at the other end reverses the process. This reduces the cost of data communications considerably.

In the mid 1960's, a man by the name of J. C. R. Licklider suggests the idea of linking computers together to share the resulting packet switched networks. This notion is turned over in many a mind, and in the spring of 1967 at the University of Michigan, the Advanced Research Projects Agency Network (ARPANET) is conceived.

In 1968, the Advanced Research Projects Agency of the U. S. Department of Defense (DoD) awards a contract to develop the ARPANET to Bolt, Beranek, and Newman Inc. (BBN), a Cambridge, Massachusetts think-tank. This network will help keep information flowing throughout America in time of war. It will also be useful in peacetime, since it will help researchers:

• share data instead of duplicating it
• minimize development times
• encourage cooperation among institutions working on different parts of the same contract

The working plan becomes this: the network will be comprised of multiple telephone lines connecting specialized hardware called "nodes" together. These nodes will accept messages from computers hooked to the node. The messages will be split into packets prior to transmission, with each packet containing a destination address and the packet's location within the original message. The node will then transmit these packets to their destinations. At the destination node, the original message will be reconstructed from the individual packets, and the message will then be sent to its ultimate destination. The network will not guarantee successful delivery of messages, so protocols must take this into account. Also, to boost reliability, packets of a single message may travel over different paths to their destination.

The year in 1969. The country is still recovering from the euphoria surrounding the climactic landing of the first man on the moon. Late in the year, the efforts of the network contract pays off as the first ARPANET node is installed at the University of California at Los Angeles. Three more nodes are installed at the University of California at Santa Barbara, Stanford Research Institute, and the University of Utah. These four machines comprise the entire ARPANET.

The 1970's
In 1970, three more nodes were added on the East coast. They were located at the Massachusetts Institute of Technology, Harvard University, and Bolt, Beranek, and Newman. This brings the total up to seven. The seven machines are known as Interface Message Processors (IMP), and the protocols used are the Network Control Protocol (NCP) and the Host-to-Host Protocol.

By 1973, the ARPANET has grown to 25 machines all over the U. S. For a while, the network is used for its original purpose, to help researchers share information on government contracts. As time goes on, its use expands to include other types of research and personal uses, most notably electronic mail. The ARPANET becomes its own entity and garners a following of researchers and scientists. It becomes a status symbol to have an account on the ARPANET. It is also expensive. Costing $250,000 a year to maintain a network connection, only big research facilities can afford to be "on the net."

The year 1973 also ushers in transatlantic communications, as England and Norway are brought on-line. The ARPANET has now stepped outside the boundaries of the U. S. At first, the milestone goes unnoticed, but years later it will prove to be a great testament to the power of networking.

As the 1970's wear on, the number of machines on the net is rapidly approaching its maximum, 256. Also, the NCP protocol is ill-designed to keep up with the amount of traffic now flowing over the network. Clearly a successor protocol to NCP is needed.

Vinton Cerf and Robert E. Kahn describe their ideas on this successor protocol in a technical paper published in 1974 for the International Networking Group (INWG). The newly renamed DARPA (Defense ARPA) contracts Cerf and Kahn to develop a new protocol.

Cerf and Kahn concentrate first on the NCP address field. An NCP packet contains the destination machine's numerical address on a one-byte field. This constrains the ARPANET to 26 machines. The design team decides that this field should be increased to 4 bytes in length. Also, this 4 byte address should be divided into a network portion and a host portion. This will allow institutions to build "mini" ARPANETs within the facility and hook every machine to the ARPANET.

The 1980's
In 1981, some new faces appear on the horizon. BITNET (Because It's Time Network) and CSNET (Computer Science Network) sprout up and hook into the ARPANET. These three networks become the major players amidst a sea of smaller ones. However the added coverage comes at a cost: the number of machines on the ARPANET has almost exceeded the limt of the NCP protocol. Fortunately, a solution is nearing.

In 1982, the final design of the new ARPANET protocol is finished and implemented. It is christened the Internet Protocol (IP). IP is one of 4 protocols that have been designed. The others work on top of IP. They are: User Datagram Protocol (UDP), Transmission Control Protocol (TCP), and Internet Control Message Protocol (ICMP). UDP provides the non-guaranteed service that was previously provided by the NCP protocol. TCP provides a more robust guaranteed delivery service, and ICMP lets machines return error conditions across the network.

Enter the year 1985. The National Science Foundation decides that easy access to massive computing power for researchers is need. To do this, the NSF creates 5 supercomputer sites across the country. This move will give researchers more access to the computing abilities that only machines like the Cray I (A large supercomputer manufactured by Cray Research, Inc.) can provide. In the past, only large corporations and weapons developers had access to these resources. The NSF installs machines in Champaign IL,. Ithaca NY, Pittsburgh PA, Princeton NJ, and San Diego CA. Since the machines need to be accessible to researchers, each machine is connected to the ARPANET.

Shortly after the supercomputers come on-line, the load introduced by the new machines exceeds the ARPANET's capacity, even with the new IP protocol. To solve the problem, NSF quickly engineers a temporary network that ties all of the NSF computer sites together. This temporary network, known as the original NSFNET backbone, removes the excess load from ARPANET. However, to remain accessible to researchers, the NSFNET ties into the ARPANET at Carnegie Mellon, which is connected to both networks.

Now, this new backbone is only considered a stopgap measure, instituted to allow researchers to access the supercomputers until a new NSFNET backbone can be deployed. The NSF notes that the supercomputers require network capacity orders of magnitude larger than the ARPANET, which runs at 56,000 bps at present. Therefore, the new NSFNET will run 20 times faster.

The year 1987 rolls around. The NSF issues requests for proposals from groups who want to establish the new NFSNET. On November 24, 1987, NSF announces that it has selected a partnership to implement the new network. The partnership consists of: MERIT Inc., a Michigan state computer network operator; IBM Corporation, a computer manufacturing giant; and MCI Inc., a long distance telephone carrier. The partnership designs and installs the new NSFNET, which starts operating during the summer of 1988. Soon afterward, the original NSFNET is taken out of service and disconnected.

In 1989, the NSFNET undergoes some relatively minor changes. Traffic flows on the NSFNET have been measured, and the resulting data is used to add and delete connections, thus maximizing the use of the network. Also, to boost reliability, each site is now given redundant connections to the network. Finally, the capacity is further increased, resulting in speeds of 1.544 Mbps (27 times faster than the original ARPANET).

The 1990's
Out with the old, in with the new, as they say. Well, as 1990 comes into view, the original ARPANET comes to an end. Over the years, the original ARPANET has been subsumed by the NSFNET and the many smaller computer networks that have sprung up and connected into it. This resulting maze of networks is collectively titled the "Internet." From this point on, less and less attention is paid to the individual network identities. No longer is it the playground of the elite and scientific. It is at this point that the actual number of machines attached to the net becomes ambiguous (see sidebar How Fast Is it Growing?). In addition, businesses are coming on-line, and the number of users is skyrocketing.

As businesses come on-line, they hit a roadblock. The NSF is enforcing an Acceptable Use Policy (AUP) on all packets that travel over the NSFNET portion. This AUP prohibits any uses related to commercial business. Since the businesses can't use the NSFNET for commercial endeavors, General Atomics, Performance Systems International, and UUNET Technologies creates a new network that bypasses NSFNET for such traffic: the Commercial Internet Exchange (CIX) network.

With that we come to the present, 1994. The NSFNET backbone now runs at 44.736 Mbps, which is 30 times faster than it ran in 1990. Also, the NSF has decided to turn over funding of the NSFNET to the private sector and lift the AUP. This has both good and bad consequences. The NSF has been funding $18-20 million of the cost of the NSFNET, so this money must be collected from new sources. However, the lifting of the AUP will encourage businesses to use and pay for the NSFNET.

Happy Birthday to I
The Internet celebrated its 25th birthday in September, 1994. So, the Internet is older than most people think. However, only since the creation of the NSFNET and the CIX network have people been given the chance to actually use the network. These events and the ever-increasing number of people using the Internet has made it "noticeable" Now that it has been noticed, some people are scurrying to find uses for what they consider a long-lost treasure.

In many ways, the Internet in the 1990's is much like the New World in 1492, after Christopher Columbus discovered it. Some people, when first made aware of the existence of the Internet, charged that it was a hoax and that it was not possible. Others gasped in awe at the massive speeds and breadth of coverage of the Internet. People are coming on-line in droves, and some are unknowingly upsetting the delicate balance that exists in the Internet. This balance concerns information and its dissemination. Though, as time goes on, you can rest assured that the Internet will overcome this much like it has surmounted previous obstacles.

Does the Internet have a future?
The present slowly turns into the past, and the future likewise becomes the present. As you sit here reading this, the future of the Internet is being planned and implemented. The NSF is removing both its funding and the AUP, while creating a new high speed backbone called the vBNS (Very High Speed Backbone Network System) that will run at 155Mbps. This new network will service researchers who need the extra speed to complete projects using the NSF supercomputers.

Not only will a new network emerge, but a new protocol for the Internet will be introduced as well (see sidebar 4,294,967,296 Isn't Big Enough!). One of the most promising specifications is IPng (Internet Protocol Next Generation). Designed by many of the same people who worked on IP, the new protocol will reflect the changes in use that the Internet has seen.

The Internet's name may change as well. At this time, the National Telecommunications and Information Administration is accepting proposals for the National Information Infrastructure (NII), which is the correct term for the "Information SuperHighway." Possibly, the Internet will form a large part of this new network that will go into every home and business.

Well, lest you think that this is all of the history behind the Internet, let me assure you that one article cannot do justice to this subject. I have attempted to describe the more shaping events, while leaving equally interesting but less influential items out. In fact, many books have been written on just parts of the history.

I believe a knowledge of the Internet's history serves a very important purpose. A knowledge of history sometimes describes the culture that has evolved. It is this way on the Internet. Too often people hop on without finding out how things operate inside the network, and they find out the hard way. As I would like to see Commodore users hop on as well, this is my attempt to smooth the bumpy road to success on the Internet. Once you've arrived, look the Commodore gang up and say "Hi."