How the Internet Was Built, and Why It Works the Way It Does

The internet's history is usually told as a list of dates. It is more useful told as a series of problems and the decisions that solved them — because those decisions still determine how it behaves when your connection is slow, when a site goes down, and when something will not load.

The problem it was built to solve

Computers in the 1960s could not talk to each other. Each was a separate machine, with its own terminals, incompatible with everything else.

Connecting them by dedicated lines did not scale. A line between every pair of machines means the number of lines grows impossibly as machines are added.

The first insight was packet switching, developed independently in the United States and Britain in the early 1960s. Instead of opening a dedicated circuit for the duration of a conversation — as telephone networks did — break the message into small packets, send each independently, and reassemble at the far end.

Why this matters today: packets can take different routes, so a failure anywhere does not break the conversation. It is also why your connection degrades gradually rather than cutting off, and why latency and bandwidth are different things.

ARPANET

In 1969 the first nodes of ARPANET were connected, funded by the United States defence research agency, linking a handful of university computers.

The first message sent was intended to be "LOGIN". The system crashed after two characters, so the first thing transmitted was "LO".

By the early 1970s it had grown to dozens of nodes, and something unplanned had become its main use: electronic mail. The @ symbol was chosen in 1971 to separate the user from the machine, for the practical reason that it appeared on keyboards and in no one's name.

A pattern worth noticing: the network's dominant use was not what it was built for. That has repeated at every stage since.

The protocols that made it a network of networks

The problem by the 1970s: several separate networks existed, each with its own rules, and they could not interconnect.

TCP/IP was the answer, developed through the 1970s and adopted by ARPANET on 1 January 1983 — a date sometimes given as the internet's birthday.

The design decision that mattered: the network itself stays simple and does not guarantee delivery. The complexity sits at the ends, in the devices. Any network that speaks the protocol can join without permission from anyone.

This is why the internet grew as it did. No central authority had to approve a new network, and no one had to redesign the middle to add a new kind of use. It is also why the internet has no owner and no off switch.

Names instead of numbers

Every machine has a numeric address. Remembering them is impractical.

The Domain Name System arrived in 1983, mapping names to numbers in a distributed hierarchy rather than one central list.

Why you should care: DNS is why "the internet is down" is frequently not down at all. When pages will not load but the connection is fine, DNS is the usual cause, and changing the DNS server often fixes it. See wireless connection problems.

The web, which is not the internet

A distinction worth keeping. The internet is the network. The web is one thing that runs on it, alongside email, video calls, file transfer and much else.

Tim Berners-Lee proposed the web in 1989 at CERN, to solve a specific problem: researchers could not easily share documents across incompatible systems.

Three pieces made it work — a way to address documents, a protocol to request them, and a language to write them in. All three still underlie every page you open.

The first website went live in 1991.

In 1993 CERN placed the web technology in the public domain, at no cost to anyone. That decision is arguably the single most consequential in the whole history, and it is the reason the web became universal rather than one of several competing commercial systems.

Opening up

Until the early 1990s, commercial use was restricted. The network was for research and education.

Those restrictions lifted in 1991, and commercial providers began connecting the public.

Mosaic, in 1993, was the first browser to display images alongside text and to run on ordinary personal computers. It turned a specialist tool into something anyone could use, and the growth from that point was extraordinary.

Search became necessary as the number of pages grew past the point where directories worked, and the search engines of the late 1990s established the model still in use.

What changed after that

Broadband replaced dial-up through the early 2000s, and the always-on connection changed what was possible more than the speed did.

Wireless networking removed the requirement to be at a desk.

Mobile made the internet something carried rather than visited. For a large part of the world's population, a phone was the first and only way they used it.

Cloud services moved computing back to remote machines — a return, in some ways, to the model that ARPANET's designers were trying to replace.

Streaming became the dominant traffic type, which is why residential connections are engineered for download far more than upload, and why video calls feel worse than the download figure suggests.

Why the history explains the present

Packets take different routes, which is why a problem can affect one site and not another, and why the route matters as much as the speed.

No central authority, which is why nobody can switch it off and why standards move slowly.

Complexity at the edges, which is why your device does the work of reassembling and retrying, and why an old device on a fast connection still performs badly.

Names layered over numbers, which is why DNS failures look like the whole internet being down.

No delivery guarantee in the network itself, which is why lost packets are retried and why latency and packet loss matter separately from bandwidth. A connection with high speed and high latency is poor for video calls and fine for downloads.

The dates worth remembering

Year	What happened
Early 1960s	Packet switching proposed
1969	First ARPANET nodes connected
1971	Email and the @ symbol
1983	TCP/IP adopted; DNS introduced
1989	The web proposed at CERN
1991	First website; commercial restrictions lifted
1993	Web technology placed in the public domain; Mosaic released

The short version

Packet switching solved the scaling problem and it is why connections degrade rather than drop.

TCP/IP made a network of networks possible without central permission, which is why it grew and why it has no owner.

DNS put names over numbers, and it is the usual cause when pages will not load on a working connection.

The web is one thing running on the internet, not the internet itself.

And CERN releasing the web technology freely in 1993 is why it became universal rather than one commercial system among several.

For broader technical standards and computing references, see NIST Information Technology Laboratory.