Egypt · 27 January 2011, 22:34 UTC

II

The question that matters is not how it was done but why it could be done, and the answer is in the structure of the Egyptian market at the time [four licences of international scope].

The country's international traffic passed through four licensed operators. Four. The other providers, of which there were dozens, bought their international connectivity from those four, so that switching off four was switching off all.

On top of that, the submarine cables that reach Egypt land at a small number of stations, and the permits to operate them are granted by the state. The concentration was not a market accident; it was the result of a licensing regime that had been designed, years earlier and for reasons of revenue and of control, so that there would be few holders.

There was an exception, and it confirms the mechanism rather than contradicting it. Noor Group, a small operator, went on working for several more days. Through it passed the connection of the Cairo stock exchange and of several financial institutions; the reading made at the time —and which nobody has contradicted with documents— is that switching it off had a cost the other three calls did not have.

The switch existed. It also had an exception clause, and the clause was money.

III

That was presented at the time as an exceptional episode of a government in trouble. Thirteen years later it is not exceptional at all.

The organisation that keeps count documented in 2023 almost three hundred deliberate shutdowns in some forty countries; it was the worst year in the series. The recorded reasons are official examinations, protests, elections, military operations and disturbances, and the first on that list is not a misprint.

India has topped the list for years, with regional shutdowns decreed by district. Iran switched off almost everything for a week in November 2019, during the protests over the price of fuel. In Burma, after the coup of 2021, there were scheduled nightly shutdowns for months, from one in the morning to nine, with the regularity of a transport timetable.

The procedure is the same in every case and requires no special technology. The licensed operators are ordered to withdraw their announcements, or to filter, or to lower the bandwidth until it is useless [the practice has a name of its own: throttling]. The order can be verbal. Often it is.

There is a feature of the procedure that it is as well not to overlook. Almost never is everything switched off: a province is switched off, or mobile access is switched off and the fixed line left, or the payment systems are left working and the rest is cut.

That is not moderation. It is that a total blackout has an economic cost that can be calculated and that somebody calculates, and what is sought is the cheapest combination that achieves the effect. There are published estimates of the daily cost of a shutdown by country, and the governments that decree them know them.

IV

What separates a country that can be switched off from one that cannot is its topology, and it is as well to say so without consolation because it works in both directions.

A country with four international operators and two landing stations is switched off with four calls, and that is arithmetic before it is politics. A country with two hundred independent autonomous systems, several internet exchange points and landings on different coasts is not switched off with calls; cables would have to be cut, and cutting cables leaves marks and takes time.

That is nobody's moral virtue. It is the result of regulatory decisions taken for reasons that had nothing to do with this: how many licences are granted, whether the dominant operator is obliged to give up access, whether there is a national exchange point or everybody buys transit abroad.

And there is a second-order effect that deserves the name. When a country has no exchange point of its own, the traffic between two of its own providers goes abroad and comes back — to London, to Marseille, to Amsterdam. Two neighbours in the same city write to each other via another continent — paying transit in both directions and passing through jurisdictions that are not theirs.

For years that was the normal situation in a good part of Africa. The regional exchange point programmes that started in the twenty-tens were set up to fix exactly that, and what they measured when they finished was not only the drop in latency: it was the drop in the bill.

V

Baran built his argument on an assumption that in his context was self-evident: that links are cut because somebody destroys them from outside.

The network he was modelling was military, the adversary was external and the damage was physical. With that assumption, distributing links is a defence, and the more there are, the better.

Egypt's blackout was not an external attack and it destroyed nothing. The cables stayed whole, the equipment stayed switched on and the staff stayed at their posts. The only thing that happened is that the licence holders stopped announcing what they knew how to reach, which is exactly what the protocol allows them to do at any moment and without giving explanations.

Against that, redundancy of links is no defence, because the links were not broken. A country can have the densest mesh in the world and still be switched off, if the ones who decide whether to announce it are four people.

The question Baran did not ask —and had no reason to, in 1964— is not how many nodes are needed for the network to survive. It is how many different people have to say yes for the network to work.

In Egypt, in January 2011, it was four.

Mubarak resigned on 11 February. The internet had come back nine days earlier, on the 2nd, at 09:29 in the morning. It took less than an hour to be whole, because restoring is faster than cutting: nobody had to be called twice.