How the internet works

The internet is not a single network owned by anyone. It is tens of thousands of separate networks, run by universities, companies, governments and internet service providers, that have agreed to exchange traffic using common rules. Each independently run network is called an autonomous system.

Those networks connect at internet exchange points, buildings where many providers plug into shared switches, and through private links between large carriers. Most traffic between continents travels through undersea fibre-optic cables, not satellites.

An old telephone call reserved a dedicated circuit for its whole duration. The internet works differently. Data is broken into small chunks called packets, each carrying the address of its destination. Packets from many conversations share the same links, and each one can take a different route.

This idea, packet switching, was developed in the 1960s by Paul Baran and Donald Davies, and it was used in ARPANET, the US research network that first connected computers in 1969. Packet switching makes efficient use of links and keeps working when parts of the network fail, because traffic simply flows around the damage.

Every device on the internet needs an Internet Protocol address. The older IPv4 system uses 32-bit addresses written as four numbers, such as 192.0.2.10, which allows about 4.3 billion addresses. That is not enough, so the newer IPv6 uses 128-bit addresses, allowing a practically unlimited number.

Routers are computers that forward packets toward their destination. Each router looks at a packet's destination address and consults a table saying which neighbour is closer. Between large networks, the Border Gateway Protocol lets autonomous systems announce which address ranges they can reach, which is how the global map of routes is built and constantly updated.

• Link layer: moves data across one physical connection, such as Wi-Fi or Ethernet.
• Internet layer: IP gets packets from one address to another across many networks, with no guarantee of delivery.
• Transport layer: TCP numbers packets, confirms they arrived, resends lost ones and puts them back in order; UDP skips those checks for speed, which suits live video and games.
• Application layer: protocols such as HTTP for the web, SMTP for email and DNS for names.

People remember names, but routers need numbers. The Domain Name System translates a name like example.org into an IP address. When your device looks up a name, it asks a resolver, which works through the hierarchy: the root servers point to the servers for .org, which point to the servers for example.org, which give the final answer.

Answers are cached for a set time so the same lookup does not have to repeat the whole journey. Because nearly every connection starts with a DNS lookup, DNS failures can make working websites seem to vanish.

When you open a website, your browser first resolves its name through DNS. It then opens a connection to the server's IP address and, for an https address, performs a TLS handshake: the server proves its identity with a certificate, and both sides agree on encryption keys so nobody in between can read or alter the traffic.

The browser then sends an HTTP request, and the server replies with the page's HTML. The browser reads it, discovers the images, stylesheets and scripts it needs, and requests those too, often from content delivery networks that keep copies of popular files in data centres close to users. All of this commonly happens in well under a second.