When you start a video call, it feels as if the internet knows exactly what you're up to. Your face, your voice, your cat strolling across the keyboard mid-sentence all arrive in near real time. Surely the network understands this is a video call and gives it special treatment?

Mostly, it doesn't. The routers passing your call along have no idea whether they're carrying your face, a spreadsheet, or a meme about spreadsheets. That isn't a flaw. It's one of the cleverest decisions in engineering history: build a network that deliberately knows almost nothing, and let the devices at the edges do the thinking. That choice is why the internet could grow into things its inventors never imagined.

Consider the old telephone system. It was a marvel of central intelligence. The network set up a dedicated circuit for every call, kept track of who was talking to whom, and handled the billing. Your phone, by contrast, was gloriously simple: a microphone, a speaker and a dial. All the brains lived in the phone company's switching centres.

The internet's builders flipped this around. In the 1970s, Vint Cerf and Bob Kahn designed the protocols that became TCP/IP so the network had one main job: move packets, small labelled chunks of data, from one address to another as best it could. Anything fancier, like putting data back in order or making sense of it, would be handled by the computers at each end. In 1984, Jerome Saltzer, David Reed and David Clark gave the idea its name: the end-to-end principle.

Think of a road system. The roads don't care whether you're driving to work, to a picnic, or to your aunt's slightly awkward birthday party. They just carry vehicles from place to place. The destination and the purpose are entirely your business. That indifference is exactly what makes roads useful for nearly everything.

Takeaway

A system that does one simple thing well can support countless things it was never designed for. Generality often comes from restraint, not ambition.

The internet makes a promise that sounds almost lazy: best-effort delivery. The network will try to get your packet where it's going. It might arrive late. It might arrive out of order. It might not arrive at all. No refunds, no apologies.

This wasn't sloppiness. It was realism. Early networks ran on leased lines and temperamental hardware, and the designers assumed links would fail, routers would crash and packets would vanish. Rather than trying to make every component perfect, they made the system tolerate imperfection. When a router goes down, traffic can find another path. When a packet goes missing, the receiving computer notices the gap, and TCP on the sending side quietly sends it again.

It's a bit like sending a long letter as a stack of numbered postcards. If postcard 14 goes astray, your friend simply asks for number 14 again. Nobody needs to redesign the postal service. Because recovery lives at the edges, the middle doesn't have to be flawless. And a network that expects failure turns out to be remarkably hard to break.

Takeaway

Systems built to survive failure often outlast systems built to prevent it. Assuming things will go wrong isn't pessimism; it's design.

Because the network doesn't inspect what it carries, it doesn't approve it either. Nobody had to ask permission to invent email, the web or online gaming. When Tim Berners-Lee created the World Wide Web at CERN around 1990, he didn't need sign-off from any network operator. He wrote software, ran it on computers connected to the internet, and the network carried it like any other data.

The same goes for nearly every app on your phone. Video calls, music streaming, multiplayer games, that chess app you keep promising to delete: none of them required the internet itself to change. Compare that with the old telephone network, where for decades even plugging in equipment the phone company hadn't approved was restricted.

So what might a 'smarter' internet have cost us? Picture a network that only carried approved applications, tuned for what its designers expected in the 1970s, such as file transfers and remote logins. It would have been efficient, tidy and completely unprepared for YouTube. Today's internet isn't perfectly dumb, of course. Firewalls and traffic management exist. But the core idea endures, and it's why your next favourite app might be getting built in someone's bedroom right now.

Takeaway

Sometimes the most valuable thing a platform can offer is simply getting out of the way. Every rule about what's allowed is also a quiet decision about what will never be invented.

The internet's genius lies less in what it does than in what it refuses to do. By keeping the middle simple and trusting the edges, its designers built something that could grow far beyond their own imaginations, and that keeps working even when parts of it don't.

Next time a video call connects in seconds, spare a thought for the routers in between, cheerfully ignorant of what they're carrying. Sometimes the wisest design choice is knowing what not to know.