ARPANET, how four computers in 1969 started everything

Field note. Every protocol your server speaks today started here. Knowing the four original computers explains why TCP/IP has the shapes it does in the parts that don't make sense.

The internet has a specific birthday. October 29, 1969. The first message was supposed to be the word "login," sent from a computer at UCLA to one at Stanford. The system crashed after the first two letters. So the first internet message was "lo."

That's not just a fun trivia. That's the actual founding moment of every piece of network technology you use. This article tells the story.

The ARPANET timeline

The setup

In the late 1960s, the U.S. Department of Defense had a problem: it had funded a lot of computer research at universities. The computers were expensive and incompatible. Researchers at one institution couldn't easily share work with researchers at another. Mailing magnetic tapes was the actual state of the art.

The Advanced Research Projects Agency (ARPA, later DARPA) wanted to fix this. The project: build a network linking ARPA-funded research computers, so they could share work, share computing time, and (less publicly stated) demonstrate American technological leadership.

The project was led by Lawrence Roberts at ARPA, with technical execution by Bolt, Beranek and Newman (BBN), a Boston-area contractor.

The radical idea: packet switching

The phone system in the 1960s was circuit-switched: when you made a call, the system reserved a path through the network for your call, end-to-end. The path was dedicated to you for the duration. Reliable, but inefficient for bursty data.

ARPANET took a different approach: packet switching. Data would be chopped into small chunks (packets), each with a destination address. Each packet would be routed independently through the network. Packets from many users could share the same link, interleaving.

This sounds obvious now. In 1969 it was contested. Phone-company engineers thought it was inefficient and unreliable. Paul Baran (RAND Corporation) had proposed the idea years earlier; the phone companies dismissed it. Donald Davies (UK NPL) had independently invented packet switching around the same time.

ARPANET would be the first major real-world demonstration of packet switching. Either the idea worked or it didn't.

The first four nodes

The first four nodes were:

1. UCLA (University of California, Los Angeles). Leonard Kleinrock's group. SIGMA 7 computer.
2. SRI (Stanford Research Institute, now SRI International). Doug Engelbart's group. SDS 940 computer.
3. UCSB (University of California, Santa Barbara). IBM 360/75.
4. Utah (University of Utah). DEC PDP-10.

Each node had an Interface Message Processor (IMP), a small computer built by BBN that handled the network protocol. The IMP talked to the local research computer and to other IMPs over leased phone lines.

By December 1969, all four nodes were connected. The network worked. The first internet was online.

The first message

October 29, 1969, around 10:30 PM Pacific. UCLA's Charley Kline tried to log into the SRI computer. He typed "L." It arrived at SRI. "O." Arrived. "G." The SRI system crashed.

Kline's lab logbook recorded the event. The crash was a memory bug at SRI, not a network failure. The first message was "lo." About an hour later, after fixing SRI, the full word "login" went through successfully.

This is the moment historians point to as the birth of the internet. It was a research milestone, not a public event. Only a handful of people knew.

Growth in the 1970s

ARPANET grew steadily through the 1970s:

• 1971: 15 nodes.
• 1973: 35 nodes (including the first international node at University College London).
• 1975: 57 nodes.
• 1981: 213 nodes.

The protocols evolved as the network grew. The original Network Control Protocol (NCP) handled host-to-host communication. As the network expanded and other networks (like the Hawaiian ALOHAnet and various satellite networks) wanted to interconnect, NCP showed limitations.

In 1974, Vint Cerf and Bob Kahn published "A Protocol for Packet Network Intercommunication." This was the blueprint for what became TCP/IP. By 1983, ARPANET would transition entirely from NCP to TCP/IP. That January 1, 1983 cutover (sometimes called "flag day") is often counted as the second birthday of the modern internet.

What ARPANET enabled

The 15 years of ARPANET produced more than connections. The network was a laboratory for inventions that shaped the internet to come:

Email. Ray Tomlinson at BBN sent the first network email in 1971. He picked the @ sign to separate user from host. The convention stuck.

File transfer. FTP was developed in the early 1970s.

Remote login. Telnet existed from the start, in primitive form.

Mailing lists. Active by the mid-1970s, used heavily by ARPANET's research community.

News and discussion. USENET emerged in 1979, separate from ARPANET initially but reaching its users.

TCP/IP itself. The protocol that runs the internet today was developed and tested on ARPANET.

The pattern: ARPANET wasn't designed as a specific application. It was a substrate. Researchers built applications on top, many of which became foundational.

The DARPA connection and the Cold War myth

A persistent myth: ARPANET was designed to survive nuclear war.

This is half-true at most. Paul Baran's packet-switching work for RAND in the early 1960s did explore communication systems resilient to nuclear strikes. ARPANET drew on this work. But ARPANET itself was built for ordinary research connectivity, not for nuclear-survival communications. Larry Roberts, who led ARPANET, has explicitly said the survivability angle was secondary.

The myth probably persisted because:

• It's a good story.
• ARPANET's resilience to component failures, even though designed for other reasons, looked like nuclear-war design after the fact.
• The DoD funding source made it easy to assume military motivation.

The truth is more banal: a research agency wanted researchers to share computers and built a network for it.

When ARPANET became "the internet"

The word "internet" started appearing in the early 1970s, originally as shorthand for "interconnection of networks." By the early 1980s, "the internet" referred specifically to the network of networks using TCP/IP, of which ARPANET was the central piece.

In 1985, the National Science Foundation began funding NSFNET, a parallel network for academic research. NSFNET and ARPANET interconnected. NSFNET grew rapidly; ARPANET grew slowly.

By 1989-1990, NSFNET had eclipsed ARPANET. ARPANET was formally shut down in February 1990. Its 21-year run was over.

The internet continued, with NSFNET as its backbone, until commercial restrictions on NSFNET were lifted in 1995 and the modern commercial internet emerged.

What it cost

The total ARPANET budget across its life was small by government standards: perhaps $20 to $30 million in 1970s dollars. Adjusted for inflation, maybe $200 million in 2026 money.

For that investment, the U.S. got:

• A working demonstration of packet switching.
• The protocol foundations (TCP/IP) that run the modern internet.
• A generation of researchers trained in network design.
• Email, FTP, remote login, mailing lists, and many other protocols.
• The substrate for the entire commercial internet.

Few research investments have returned more.

What the founders missed

Reading retrospectives by Kleinrock, Cerf, and others, a few themes emerge about what surprised them:

The commercial explosion. Nobody in the 1970s expected the network to become a consumer product.

Email. Predicted to be a minor feature; became the dominant use.

Spam. First spam email sent in 1978. Some people noticed it was annoying. Nobody predicted how much.

Privacy. Designed-in privacy mechanisms were minimal because the early network was a trusted research community. The retrofitting of privacy and security has been a 30-year project.

Scale. They thought maybe thousands of computers, eventually. They had no model for billions.

The technology was sound. The applications and scale exceeded any reasonable forecast.

The cultural legacy

ARPANET's culture shaped what came after:

Open protocols. RFCs (Requests for Comments) started in 1969 as informal notes circulating among researchers. The "anyone can publish, debate happens publicly, useful ones become standards" model persists in the IETF today.

Trust between operators. ARPANET researchers knew each other. BGP and the early internet inherited that trust, which mostly still works.

Permissionless innovation. Nobody asked permission to build email on top of ARPANET. The network was a substrate; you could build whatever. This still feels like a foundational property.

Academic-commercial blend. ARPANET was government-funded but operated like academia. The transition to commercial use in the 1990s preserved a lot of academic norms even as money entered.

A small note on UCLA

The first ARPANET node was at UCLA. The room (Boelter Hall 3420) still exists. There's a plaque. The original IMP is preserved in a small museum on campus.

Wayfinding signs in the building point to "Birthplace of the Internet." The location has become a minor pilgrimage destination for people who care about this history.

It's a small office. Linoleum floors. A few desks. The internet started here.

Conclusion

ARPANET was the seed of everything we now call the internet. Four computers in 1969 became 213 by 1981 became the global commercial internet by the late 1990s. The protocols, the culture, and the design choices made in those first years still shape every interaction on the network today.

Most users don't know any of this. They use the internet daily and the history is invisible. Knowing it adds depth: you're using a 55-year-old idea, refined and scaled, but recognizably the same network those four nodes started.

Coming up

Next: how TCP/IP beat the other protocol candidates and became universal. The internet's success isn't just ARPANET's; it's the protocol that came out of it.