How the Internet Actually Works: Packets, Cables and the Address Book of the World

Ask a child where the internet is and they will say “the cloud,” which is the single most successful piece of misdirection in modern technology. The internet is not in the sky. It is overwhelmingly at the bottom of the ocean. Hundreds of fibre-optic cables, many no thicker than a garden hose, lie on the seabed carrying the great majority of all international data traffic. Satellites handle a small fraction. The cloud is a warehouse full of computers in an industrial park, usually somewhere with cheap electricity.

Getting this right matters more than it sounds. A child who thinks the internet is a mysterious ambient substance can’t reason about why it’s slow, why it breaks, why some things are private and others aren’t, or why a country can switch it off. A child who knows it’s cables, addresses and copies of messages travelling in numbered pieces can reason about all of that.

The Core Idea: Chop It Up

Packet (definition): A small chunk of data - typically around 1,500 bytes at most - wrapped in a header containing the sender’s address, the destination address, a sequence number and error-checking information. Everything sent over the internet is broken into packets.

This is the founding decision of the whole system and it’s genuinely counterintuitive, so it’s worth explaining by contrast.

The old telephone network used circuit switching. Making a call reserved a physical path from you to the other person for the entire call. Dedicated, reliable, and enormously wasteful - the line sat idle during every pause in conversation, unavailable to anyone else.

The internet uses packet switching. Your message is chopped into pieces, each labelled with where it’s going. Each piece is passed from router to router independently, and each router decides where to send it next based on current conditions. The pieces may take different routes. They may arrive out of order. The receiving computer reassembles them using the sequence numbers, and asks for any missing pieces again.

The analogy that works with kids: instead of hiring an entire train to carry one letter, you write the letter on a hundred postcards, number them, and post them all. They travel by whatever routes are quickest. The recipient sorts them by number. Anything lost gets requested again.

Two consequences follow immediately, and they explain a lot of everyday experience:

• The network survives damage. If a route fails, routers simply send subsequent packets a different way. This resilience was a founding design goal.
• Nothing is guaranteed to arrive. The base layer makes no promises. Reliability is added on top, by protocols that check and re-request.

TCP and UDP: the two personalities

Worth naming, because the difference explains video calls.

TCP checks everything. It confirms each packet arrived, requests missing ones, and puts them in order before handing them over. Web pages, emails and file downloads use it, because a page with a missing chunk is broken.

UDP doesn’t check. It fires packets and moves on. Live video and voice calls use it, because in a live call a packet that arrives late is useless anyway - you’d rather have a momentary glitch than a growing delay. That’s why a bad video call goes blocky rather than pausing to wait. The system is deliberately choosing to lose data instead of falling behind.

Addresses: Every Device Has a Number

IP address (definition): A numerical label identifying a device on a network, used to route packets to it.

The old format, IPv4, looks like 192.168.1.14 - four numbers, each 0 to 255, because each is one byte. That gives about 4.3 billion possible addresses, which seemed limitless in the 1980s and ran out in practice. IPv6 uses much longer addresses and provides a number so large it has no useful comparison.

Most home devices don’t have a public address at all. Your router has one, and everything inside your house shares it through NAT (Network Address Translation), which is why home addresses so often start with 192.168. The router keeps a table of which internal device asked for what, and directs replies accordingly - a receptionist taking calls for a whole building on one number.

DNS: The Address Book Nobody Sees

Nobody types IP addresses. We type names. Something has to translate.

DNS (Domain Name System) is that translation layer, and it’s a distributed database spread across the entire world with no single copy.

When a device needs to find a name it hasn’t seen recently, the lookup walks a hierarchy: root servers know who handles .com, those servers know who handles a given domain, and that domain’s own nameservers hold the actual address. The answer is then cached - by the device, the router and the internet provider - so the next lookup is instant.

The caching is why DNS changes take hours to spread, and why “have you tried it on a different network?” sometimes produces a different result.

DNS is also the internet’s most common point of failure and control. Several of the largest internet outages in recent years were DNS problems rather than cables or servers. And when a government blocks a website, the crudest method is instructing providers to give the wrong DNS answer - the site is still there, but nobody can find its address. That makes DNS a good concrete entry point into a conversation about censorship and infrastructure control with an older child.

The Physical Layer: It’s Cables

This is the part that most changes a child’s mental model.

The great majority of international internet traffic travels through fibre-optic cables on the seabed. Data moves as pulses of light through glass strands, with repeaters spaced along the route to boost the signal. The cables are laid by specialist ships, and they break regularly - ship anchors, fishing gear, undersea landslides - which is why repair vessels exist as a permanent industry.

Satellite internet, including low-orbit constellations, is genuinely important for places cables don’t reach, and for ships, aircraft and remote regions. But it carries a small share of total traffic, and it’s slower in one specific way a child can calculate: a geostationary satellite sits about 36,000 km up, so a signal round trip is at least a quarter of a second before any processing. Low-orbit constellations fly much closer, which is precisely why they were built.

Between the cables and your device: your data goes from your home over copper or fibre to your provider, through their network to an internet exchange point - a building where many networks physically interconnect - and onward. A handful of these buildings, in cities like Frankfurt, Amsterdam and London, carry an extraordinary share of Europe’s traffic.

The one-second journey

Loading a page involves, roughly: a DNS lookup to find the address; a connection handshake with the server; a request; the server assembling a response, often by querying its own databases; and the response returning as packets over the same infrastructure - frequently from a CDN (content delivery network) copy stored in a data centre near you rather than from the original server. Popular content is deliberately cached close to users, which is why a video from a distant country starts instantly.

Why This Matters for Privacy

Packets pass through many networks, and each one can see the headers - who sent it, where it’s going, how big it is. Encryption hides the contents but not the envelope.

So with HTTPS, your provider can see which site you connected to, when, and how much data moved, but not what you read or typed. Unless DNS is encrypted too, they can also see every name you looked up. This is a much sharper way to explain online privacy to a teenager than vague warnings: the postcard analogy again, where the message is in code but the address is written plainly on the outside, and everyone who handles it can read that.

Building It: A Network With No Internet

The clearest way to understand what the internet adds is to build something that works without it.

The CircuitMess Chatter 2.0 ($149) makes a pair of handheld communicators that message each other directly over LoRa radio, encrypted, with no router, no provider, no DNS, no account and no internet at all.

The educational value is in the contrast. The child immediately discovers what the internet was providing: reach beyond a couple of kilometres, the ability to contact someone whose device you’ve never paired with, and the ability to find someone by name rather than by prearrangement. Every one of those is a service built on top of the physical layer, and losing them makes them visible. Our guide to IoT projects for kids covers what happens when small devices do join the network.

Six Experiments to Run at Home

1. Traceroute. On any computer, run tracert (Windows) or traceroute (Mac/Linux) followed by a website name. You’ll see every router the packets passed through, with timings. Try a local site and a distant one.
2. Ping the world. ping a server in another continent and note the round-trip time. Divide the distance by the speed of light and compare - the gap is switching and routing overhead.
3. Look up an address. nslookup a domain name and see the IP behind it. Try a big site twice from different networks; the answers often differ, because a CDN is sending you to a nearby copy.
4. Find the cables. Public undersea cable maps are freely available. Find which ones land nearest your country.
5. Break a video call deliberately. Start a call, then start a large download on the same connection. Watch the video degrade rather than pause - UDP choosing speed over completeness.
6. Count the hops. Compare traceroute output on mobile data versus home broadband to the same destination.

Where This Leads

Networking underlies essentially every technical career and quite a few non-technical ones. But the immediate benefit is diagnostic: a child who understands the chain from device to router to provider to exchange to server can actually reason about why something is broken, rather than restarting things at random.

There’s a civic dimension too. Infrastructure that most people treat as ambient and ownerless is in fact owned, physically located, and subject to national law. Knowing where the cables land and who controls DNS is the beginning of understanding a large category of modern political argument.

Frequently Asked Questions

How do you explain the internet to a child?

It’s a worldwide network of cables and computers that pass messages along. Your message is chopped into numbered pieces called packets, each labelled with its destination. The pieces travel separately, possibly by different routes, and the receiving computer puts them back in order and asks for any that went missing.

Is the internet in the cloud?

No. “The cloud” means someone else’s computers in a data centre - a physical building with servers, power and cooling. The internet itself is mostly fibre-optic cable, including hundreds of cables lying on the ocean floor that carry the great majority of international traffic.

What does DNS do?

It translates names people can remember into the numerical addresses machines use to route traffic. It’s a distributed system with no single copy, and answers are cached at several levels, which is why changes take time to spread and why DNS problems cause some of the largest internet outages.

Why does video call quality get blocky instead of pausing?

Live video uses a protocol that doesn’t re-request lost packets. In a live call, data that arrives late is useless, so the system deliberately drops it and carries on rather than pausing to wait. File downloads use a different protocol that does check and re-request, which is why they slow down instead of corrupting.

Can someone see what I do online?

Encryption hides the contents of your traffic but not the addressing information. Networks handling your packets can generally see which sites you connected to, when, and how much data moved - just not what you read or wrote. It’s like a coded letter with the address written plainly on the envelope.

What age can a kid understand this?

The packets-and-postcards model works from about 9. IP addresses, DNS and the physical cable infrastructure suit 11 to 13. TCP versus UDP, CDNs and the privacy implications of packet headers are good material for 14 and up, and traceroute makes an excellent hands-on lesson at any of those ages.

The Bottom Line

The internet is cables, addresses and numbered copies of messages taking whatever route is quickest. Replace the cloud metaphor with that and a child can suddenly reason about slowness, outages, privacy and control. Then have them build something that works without any of it - a direct radio link like the Chatter 2.0 from CircuitMess -e because nothing reveals what a system provides like doing without it.