What "the internet" actually is: networks of networks, no one in charge

Field note. Knowing 'the internet' isn't one thing, just a federation of independent networks, is the difference between blaming your host and actually finding what broke between you and your players.

The internet is not a thing. It's an agreement.

That's the strange and important fact at the center of how it works. There is no single computer called "the internet." No single company owns it. No single country runs it. It is a collection of independent networks that have agreed to exchange data using a common set of rules.

This article explains what that actually means.

"Inter-network"

The word "internet" comes from "inter-network": a network of networks. In the late 1960s, several research institutions in the United States started building their own internal networks. Each network worked fine on its own. The interesting question was: could these networks talk to each other?

The answer required two innovations:

1. A common protocol that all the networks could speak when communicating with each other, even if they used different protocols internally. This became TCP/IP in the late 1970s.
2. A way to route data across multiple networks, with each network handling the part of the journey within its territory. This is what IP (Internet Protocol) does.

Once those two pieces existed, any network that adopted them could join the larger conversation. By the 1980s, dozens of networks had joined. By the 1990s, thousands. Today, roughly 75,000 distinct networks participate in the global internet.

What "no one is in charge" actually means

People sometimes say "no one is in charge of the internet" and it sounds vaguely conspiratorial. It's actually true, in a precise and technical way.

What's true:

• No single entity controls the routing of internet traffic.
• No single entity owns the cables and equipment.
• No single entity can shut the whole thing off.
• No single entity issues "internet licenses" to use it.

What's also true:

• There are international bodies that coordinate technical standards (IETF, IANA, ICANN).
• There are organizations that allocate IP address blocks to networks (the regional internet registries).
• There are organizations that manage the domain name system (ICANN and root server operators).
• Governments regulate what their own networks can and can't do.

The coordination happens at the protocol layer. The routing, ownership, and operation happen network-by-network.

This works because the internet isn't a hierarchy. It's a mesh.

The autonomous system

The building block of this mesh is called an autonomous system (AS). An AS is a network or group of networks under a single administrative control: a single internet provider, a single university, a single big company.

Each AS gets a unique number from one of the regional registries (ARIN for North America, RIPE for Europe, APNIC for Asia-Pacific, LACNIC for Latin America, AFRINIC for Africa). There are about 110,000 AS numbers issued globally, of which ~75,000 are actively visible on the internet.

AS numbers identify networks the way IP addresses identify computers. Every connection you make traverses multiple ASes, each forwarding traffic to the next.

Examples of ASes:

• Comcast is AS7922.
• Google is AS15169.
• Cloudflare is AS13335.
• A small regional ISP might have an AS number like AS399999.

We'll dig into ASes and the protocol that connects them (BGP) in a later article. For now: think of the internet as 75,000 cooperating but independent organizations, each running their own network.

How networks decide to talk

When two ASes want to exchange traffic, they have to physically connect and agree on terms. There are roughly three patterns:

Peering. Two ASes connect directly and exchange traffic for free. Usually because they exchange roughly equal amounts. Big content providers (Google, Netflix) peer with big eyeball networks (Comcast, Verizon) because both benefit.

Transit. A smaller AS pays a larger AS to carry its traffic to the rest of the internet. The smaller AS is buying access; the larger AS is selling it. This is how most ISPs in the world get connected.

IXP membership. Multiple ASes connect to a shared "internet exchange point," a neutral facility (often a specific building) where any member can peer with any other. Lower cost than bilateral peering at scale.

The picture: a smaller network buys transit from a larger network, which has settlement-free peering with other large networks, which serve smaller networks that buy from them. Traffic finds a path through this graph.

If your home internet connection is with a midsize ISP, and you load a video from a major streaming service, the packets probably flow:

You → your ISP → your ISP's transit provider → the streaming service's network

Or, often more efficiently:

You → your ISP → IXP → the streaming service's network

The exact path depends on the contractual relationships between the ASes involved. Neither you nor the streaming service controls the path. The ASes do, dynamically, using BGP.

Why the agreement works

Here's the magic. None of this is enforced by treaty. There's no global authority. So why does it keep working?

A few reasons:

Mutual self-interest. Each network wants its users to reach the rest of the internet. The way to achieve that is to interconnect. If you don't, your users will go to a network that does.

Shared standards. The IETF publishes RFCs (Requests for Comments) that document how protocols should work. Anyone can write an RFC; the IETF community evaluates and standardizes them. Once standardized, networks implement them because not implementing them means breaking compatibility with everyone else.

Coordination at the address layer. IP addresses and AS numbers are allocated through a clear hierarchy (IANA delegates to RIRs, which delegate to networks). You can't just claim 1.0.0.0/8; you have to be assigned it. This prevents chaos.

Operational trust. Network operators know each other. There are mailing lists (NANOG, RIPE-list, etc.) where issues get discussed. The world of "people who can break the internet" is surprisingly small and largely cooperative.

This trust occasionally gets abused (BGP hijacks, route leaks). When it does, the rest of the network detects and routes around it, often within hours.

The "core" of the internet

You may hear references to the "internet backbone." It's a useful image but slightly misleading.

There isn't one backbone. There are several dozen very large networks (Tier 1 networks) that operate globally and exchange traffic with each other for free. Examples include Lumen (formerly CenturyLink/Level 3), Cogent, NTT, Telecom Italia Sparkle, GTT, Tata, and a handful of others.

These Tier 1s sell transit to smaller networks. They peer with each other in many cities worldwide. Together they form the connective tissue that lets any user on any network reach any other user on any other network.

If all Tier 1s went down simultaneously, the internet would fragment. Local networks would still work; cross-continent traffic might not. This has never happened. The Tier 1 networks are diverse enough that simultaneous failure is implausible.

What this means for you

For a user of the internet, this matters in a few practical ways:

Your experience depends on which networks are between you and what you're reaching. A poorly peered ISP can give you worse latency to a specific service than a well-peered one, even at the same advertised speed.

The "path" your traffic takes can change. If a major link goes down, BGP reroutes. You might not notice; you might see a brief spike in latency.

You can sometimes choose a better path. Using a VPN, a proxy, or a smarter DNS provider can shift your traffic to use better peering arrangements.

Outages are usually local. When a service is down for you but up for someone else, the problem is usually in the specific ASes between you and the service, not in the service itself.

A simple mental model

When you load a webpage:

1. Your computer asks DNS for the server's IP.
2. Your OS makes a TCP connection to that IP on port 443 (HTTPS).
3. The packets travel from your network card, to your router, to your ISP, through your ISP's transit or peering to the next AS, eventually to the destination AS.
4. At each hop, a router looks at the destination IP, consults its routing table, and forwards the packet to the next hop.
5. The destination server receives the request and sends a response, which travels back through some path (possibly different from the outbound).

The whole journey, for a typical interaction, takes 20 to 100 milliseconds. The packet has been handled by dozens of routers across multiple ASes, none of which need to talk to a central authority to do their job.

This is the internet. A coordination achievement disguised as a technology.

Coming up

In the next articles we'll get more specific. How exactly do computers find each other (IP addresses). How they talk to specific programs on each other (ports). How they find names instead of numbers (DNS). And how a network of networks decides which way to send each packet (routing).

But the high-level picture is now in place. Hold it. Everything else fits inside it.