Why a research network was built
ARPANET was funded by the Advanced Research Projects Agency, a US defense research body. One practical motivation was sharing expensive computers: in the 1960s, research groups at different universities each wanted access to machines that only a few sites could afford. A network would let a researcher use a distant computer’s resources without traveling to it.
A common story is that ARPANET was designed to survive a nuclear attack. The reality is more mixed. Some of the underlying ideas about resilient communications came from military-funded work, but the ARPANET project itself is generally described as focused on resource sharing among researchers. It is safest to say the two threads overlapped rather than that one explains everything.
Packet switching
The telephone network of the time worked by reserving a circuit for the length of a call. Computer traffic does not fit that pattern well, because it comes in short bursts with long silences between them. Reserving a line for all that idle time is wasteful.
Packet switching handles this by breaking a message into small pieces. Each packet carries addressing information and travels separately, and the receiving computer puts the pieces back in order. Several people contributed to the idea. Paul Baran worked on distributed communications designs in the US, Donald Davies at the UK’s National Physical Laboratory developed similar thinking and is credited with the term “packet”, and Leonard Kleinrock produced early mathematical work on queues in networks. Historians still discuss how much credit each deserves, and it is fair to say the idea emerged from several people in parallel.
Because packets can take different routes, the network can keep working when one link is busy or fails. That flexibility is one reason the approach scaled so well.
The first ARPANET links
ARPANET used dedicated small computers called Interface Message Processors, or IMPs, as intermediaries. Each site’s main computer connected to an IMP, and the IMPs talked to each other over leased telephone lines. The contractor Bolt Beranek and Newman built them.
The first connection was between the University of California, Los Angeles and the Stanford Research Institute in October 1969. The often-told detail is that the first attempted message, the word “login”, crashed partway through, with only the first letters getting across. Within the same year, two more sites joined, and the network grew to a few dozen nodes in the early 1970s.
Email and early applications
Remote login and file transfer were the first intended uses. Email, introduced in 1971 by Ray Tomlinson, was not part of the original plan but became the most popular service. Mailing lists followed, giving researchers a way to discuss topics with groups across the network. Much of what is familiar now, such as messages with a sender, a subject and a recipient, started here.
The problem of many networks
By the mid-1970s, ARPANET was no longer the only packet network. Radio-based and satellite-based networks were being developed, and each used different rules. Connecting them required a shared set of agreements that sat above any single network.
Vint Cerf and Bob Kahn described such a design in 1974. Over several revisions it split into two roles: IP, which handles addressing and routing packets between networks, and TCP, which handles reliable delivery, such as noticing lost packets and requesting them again. Together they became TCP/IP. Today the idea of an IP address is familiar to anyone who has looked at a home router, and it traces directly to this work.
ARPANET officially adopted TCP/IP on 1 January 1983. This is why many accounts treat that date as the point when a collection of networks became the internet. Around the same time, the domain name system was proposed so that people could use names in place of numeric addresses.
NSFNET and wider academic access
Not every university could get an ARPANET connection, which was tied to specific defense-funded research. In the 1980s, the National Science Foundation funded other networks and then a national backbone known as NSFNET, linking supercomputing centers and regional academic networks. NSFNET adopted TCP/IP, so it joined the wider internet smoothly, and traffic grew quickly.
ARPANET itself was shut down in 1990, having by then been overtaken in speed and scale by newer networks. Its protocols and habits lived on.
Opening to commercial use
NSFNET operated under an acceptable-use policy that limited traffic to research and education. As private companies started offering their own network services, pressure grew to allow commercial traffic. The restrictions were relaxed in the early 1990s, and commercial providers began connecting customers and exchanging traffic with each other directly.
In 1995, the NSFNET backbone was retired and replaced by a structure in which commercial providers carried the main traffic. Around the same time, the web and graphical browsers made the network attractive to people who had never used a command line. Dial-up services let households connect over a phone line and modem, and the number of users and websites began rising quickly.
What carried over
Several choices made in this period still shape the internet:
- Open specifications. The protocols were published, so anyone could build compatible software.
- A simple core. The network moves packets and leaves most intelligence to the computers at the edges.
- Loose cooperation. No single organization runs the whole thing; networks agree to exchange traffic.
These choices helped the network grow, but they also left gaps. Early designs gave little thought to security or identity, which is one reason problems such as spam and fraud had to be handled later and why advice about online safety is so necessary today.