The Internet was not invented in a day or by a single person. It is the result of a series of ideas and projects that followed one another for more than a decade, starting at the end of the 1960s. Retracing this story in order helps explain why the network works the way it does and dispels some very common myths.
The idea of packet switching
Traditional telephone networks worked by circuit switching: a call set up a dedicated connection between two points, which stayed occupied for the whole conversation, even during silences. For computer traffic, made of short bursts of data separated by long pauses, this was wasteful.
In the 1960s two researchers independently arrived at a different solution:
- Paul Baran, at the RAND Corporation in the United States, was studying communication networks able to withstand the loss of some nodes, and proposed dividing messages into blocks routed separately.
- Donald Davies, at Britain's National Physical Laboratory, developed a similar idea and coined the term packet.
In packet switching each message is divided into small pieces, each carrying the destination address. Packets travel through the network, even by different routes, and are reassembled on arrival. The same link can thus be shared by many communications at once.
ARPANET
ARPA, the US Department of Defense's Advanced Research Projects Agency, funded many computing research centres. Each had different, mutually incompatible computers. Connecting them would make it possible to share expensive resources.
The network project was entrusted to Larry Roberts. Building the switching nodes, small computers called IMPs (Interface Message Processors), was contracted to the company BBN. Each centre connected its main computer to its IMP, and the IMPs were linked to one another by leased telephone lines.
The first message
On 29 October 1969 a student in Los Angeles, Charley Kline, tried to connect from his laboratory's computer to the one at the Stanford Research Institute, hundreds of kilometres away. He was supposed to type the word LOGIN. After the letters L and O the remote system crashed. The first message in ARPANET's history was therefore "LO". Shortly afterwards the connection was restored and the login succeeded.
By the end of 1969 the network had four nodes: Los Angeles, the Stanford Research Institute, Santa Barbara and Utah.
A myth to correct
You often read that ARPANET was built to survive a nuclear war. That is not correct. Baran's studies were indeed motivated by the resilience of military communications, but ARPANET's stated purpose was sharing computing resources among research centres. The two stories intertwine because ARPANET adopted packet switching, but the goals were different.
The problem of different networks
In the early 1970s ARPANET ran on a protocol called NCP. Meanwhile other packet networks were appearing: radio networks, satellite networks, networks in other countries. Each with its own rules. How could they be made to talk to each other?
Cerf and Kahn
Vint Cerf and Bob Kahn tackled this problem and in 1974 published a paper entitled A Protocol for Packet Network Intercommunication. The underlying idea was revolutionary in its simplicity:
- each network stays as it is, with its own technology;
- devices called gateways, which we would now call routers, connect the networks to one another;
- a common protocol, run by the computers at the ends, makes sure the data arrive complete and in order.
The network in the middle was to be simple and limited to carrying packets; the intelligence lay at the edges. This principle, often called end-to-end, still underlies the Internet.
Splitting TCP and IP
The original protocol performed all functions together. At the end of the 1970s it was split into two layers:
| Protocol | Task |
|---|---|
| IP | Addressing and routing individual packets between networks |
| TCP | Ensuring reliable, ordered delivery between two applications |
The split also made it possible to create different protocols on top of IP, such as UDP, suited to applications that prefer speed to reliability.
1 January 1983
The date many regard as the true birth of the Internet is 1 January 1983. On that day ARPANET abandoned NCP and officially switched to TCP/IP. From that moment ARPANET became one network within a set of interconnected networks: an internet, precisely.
In the same period the DNS was introduced, replacing manually maintained name tables with a distributed, hierarchical system.
After ARPANET
In the following years the network grew mainly through research networks. ARPANET, by then obsolete, was decommissioned in 1990.
Why this history matters
Many features of the Internet that cause problems today stem from its original context. The network was designed among institutions that trusted one another: security was not an initial requirement. Protocols such as email and name resolution were born without authentication and were reinforced later, with subsequent additions. Understanding this origin helps explain why network security needs so many additional layers.
In short
- Packet switching was conceived independently by Paul Baran and Donald Davies, who coined the term "packet".
- ARPANET's first message, on 29 October 1969, was "LO", because the system crashed while LOGIN was being typed.
- In 1974 Cerf and Kahn described the protocol for interconnecting different networks, later split into TCP and IP.
- On 1 January 1983 ARPANET switched to TCP/IP, a date regarded as the birth of the Internet.
