There Is No Head Office, And That Is The Whole Design

The internet is not a thing that somebody built and now operates. It is an arrangement between many independent networks that each chose to connect, and it holds together for reasons that are worth understanding properly.

Ask who runs the internet and the honest answer is that nobody does, not in the way anybody runs a railway or an airline or a water utility. There is no head office, no chief executive of the network, no single organisation that can switch it off or decide what it carries. What exists instead is a very large number of separate networks, each owned and operated by somebody different, each making its own decisions about equipment and staffing and customers, and each choosing to exchange traffic with some of the others. The internet is the name we give to the result of all those choices. It is a pattern rather than a possession, and once that is understood, a great many otherwise puzzling things about it become clear.

This surprises people because almost every other system of comparable scale has an owner. Electricity grids have operators. Postal systems have national administrators. Aviation has authorities that license aircraft and control airspace. The instinct to look for the equivalent body here is entirely reasonable, and the fact that the search comes up empty is not an oversight or a gap waiting to be filled. It is the arrangement working as intended. The networks that make up the internet remain independent precisely because no central authority was ever placed above them, and they cooperate because cooperating serves each of them, not because anybody instructed them to.

That does not mean nothing is coordinated. A small number of things genuinely must be agreed globally, and there are arrangements to agree them. Names have to be unique, or the same name would lead different people to different places. Addresses have to be handed out rather than chosen freely, or two networks would end up claiming the same one. These necessities have produced institutions, and those institutions have real responsibilities. But the striking thing about them is how narrow their remit is. They coordinate the minimum required for the whole to function and leave everything else to the operators themselves.

The consequence is a system with unusual properties. It can grow at the edges without permission from the centre, because there is no centre to ask. It can carry applications nobody anticipated, because the networks in the middle were never designed around any particular use. It is also remarkably hard to fix when something goes wrong at scale, because fixing requires persuading many independent parties rather than issuing an instruction. The same absence of an owner that makes the internet adaptable also makes it slow to change direction, and both facts follow from exactly the same feature.

This is not a claim that the arrangement is beyond criticism. Plenty of thoughtful people argue that a system this important should have clearer lines of accountability, and that argument deserves a fair hearing rather than dismissal. Others argue with equal seriousness that concentrating authority over a global network in any single place would create risks worse than the ones it solved. Both positions rest on real concerns. What both sides tend to agree on is the factual starting point: that the thing being argued about was not designed by anybody as a whole, and does not have an owner to appeal to.

This issue sets out to explain the arrangement rather than to defend or attack it. We look at what the voluntary agreements between networks actually consist of, at the narrow set of functions that genuinely require central coordination, at how the technical standards everyone follows come to be written and why anyone follows them, and at the pressures now bearing on all of it. The aim throughout is that a reader who finishes should be able to explain to somebody else why a network nobody runs does not simply fall apart, and what would have to change for that to stop being true.