The Network That Cannot Say Who Owns It

We are told this is a question of who holds the switch: President Zelensky’s government, officials in Washington, or Elon Musk. Put that way, it is a custody fight, and whoever wins, someone else loses. But state it as a design problem. The need is reliable command communication for Ukrainian forces across a contested front. The resources are thousands of satellites, small flat terminals that a soldier can carry in a backpack, and a ground segment run by one company. The real constraint is not bandwidth or orbit or terminals. It is that the system’s authority has a single point of failure, and nobody can say for certain where that point is. Asked that way, a different question appears: what arrangement makes the switch matter less?

The comedy writes itself, and it tells you something. Here is the most sophisticated communications system ever put over a battlefield. It can carry a video call from a muddy dugout to a ministry hundreds of kilometres away with a lag you would not notice. And it cannot carry one plain sentence: this is who decides. American officials give one account and Musk gives another. The machine moves gigabits and stalls on a single fact. That is not a technical failure. It is the signature of a system designed with great care for the payload and almost no care for its governance.

Widen the boundary and the waste shows. An engineer budgeting a bridge counts the load the bridge must carry, and also the load of uncertainty: wind, settlement, fatigue. Ambiguity about control is that kind of load on a military network, and it is paid continuously. A commander who does not know whether a sector’s coverage could be restricted for reasons decided elsewhere has to plan for that loss every time. So he keeps backup radios and couriers, and he hesitates over operations that depend on the link. The ambiguity therefore costs something even on days when the switch is never touched. It taxes every decision, in a currency no ledger records.

Picture the signals sergeant who crouches beside the dish on a sandbagged parapet and watches the connection icon turn. Is it jamming, terrain, weather, or a policy written in another hemisphere? He cannot know. The terminal at his elbow cannot tell him. That is the design brief stated as a human being.

The strongest objection deserves its full weight. Starlink works. Nothing else on offer matches its density, its price per terminal, or its speed of deployment. Splitting the load across slower or costlier systems would buy resilience with real capability, and every rival is less finished. All of that is true. Concentration is also efficient right up until the moment it is not, and redundancy is not waste when the failure it guards against is total. A single bearing is lighter than two, and also a fine way to lose the whole machine.

So the arrangement that does more with less is not a bigger constellation. It is plurality of paths. Terminals and routing should be able to fail over among more than one provider: Eutelsat’s OneWeb constellation already flies, and Europe’s IRIS2 programme is meant to follow. Short-hop radio meshes would carry local traffic between units without touching orbit at all, and buried fibre would serve the rear. Each layer is individually less capable. Together they make the network indifferent to any one owner’s hand. The material cost of this is modest next to the cost of an army that hesitates. The tetrahedron is stable not because any one strut is strong but because no single strut is load-bearing alone; communications can be framed the same way. That is the one structural lesson worth borrowing.

Design cannot settle the rest, and I will not pretend otherwise. Whether a private company may shape the operating envelope of a war is a question of law, contract, and state power. That fight runs between Kyiv, Washington, and a corporation’s leadership, and no clever topology resolves it. Whatever arrangement those parties reach, it should be published, written, and binding. An unknown answer is worse than an unwelcome one, because one can at least design around a known constraint.

The goal is not to win the switch. The goal is a network in which the switch becomes a minor part, so that when the sergeant’s icon stops turning, he simply flips to the next path, and the war does not even notice the outage.