Middle East cable corridors

Five statements about the model; the mapping to the region is a reconstruction R.

• Count corridors, not cables. Ten cables through one corridor are one route against the loss of that corridor. Treated as ten independent links (each failing with $B$'s probability, independently of the others and of $E$, so that the corridor's shared exposure to Egypt disappears as well), they would make connection look $99.6\%$ assured; keeping $E$ shared and splitting only $B$'s risk among the cables still gives $94.7\%$. Counted as the corridor they share, the probability is exactly $\frac{19}{25}=0.76$, the product of the two chokepoints' survival D [Proposition 9.13 (p. 378)(b)].

• One chokepoint can take two thirds. Losing either $E$ or $B$ cuts capacity from $18$ to $6$; two chokepoints, chosen well, leave $3$ (Table 9.1 (p. 357)). The worst pairs are not the two famous straits but a strait and a land crossing shared by both bypasses, $\{B,S\}$, or Egypt and the Levant, $\{E,L\}$: the routes designed to avoid one chokepoint concentrate on another D [Proposition 9.13 (p. 378)(a)].

• Two different robustness counts. At most four routes are pairwise chokepoint-disjoint (R1, R4, R5, R6), and four chokepoints, such as $B$, $C$, $N$, $S$, together cross every route; the two certificates prove that both numbers are $4$ for this table. The equality is a fact about this route table, checked by enumeration: for an arbitrary list of routes given as sets of chokepoints it can fail, and Menger's theorem applies only when the routes are all the paths of a graph. Four is the number of losses needed to stop everything; to break the service threshold, one loss ($B$ or $E$) is enough D [Proposition 9.13 (p. 378)(a)].

• The same capacity buys different resilience. Adding four units of capacity raises the probability of staying connected from $0.76$ by nothing if the units go into the same corridor (R1), to $0.79$ for the Egypt bypass that still crosses $B$ (R2), to $0.90$ for the bypass of $B$ that still crosses $E$ and $H$ (R3), and to $0.93$ for any route that shares no chokepoint with R1 (R4–R6) (Figure 9.6 (p. 357)). Sharing a chokepoint is not fatal here: R3, which shares $E$ with the corridor, still delivers about $81\%$ of the best gain; what it loses is the probability that $E$ fails D [from Proposition 9.13 (p. 378)(c)]. Under a regional shock that removes the Red Sea and Hormuz together the ordering is the same at lower levels ($0.68$ to $0.84$) D [Proposition 9.13 (p. 378)(c)].

• Capacity and resilience are different purchases. The same four units of capacity add nothing to the probability of service if placed on the corridor (R1) and the most if placed on a route that shares none of its chokepoints (R4, R5, R6). A planning document that reports only terabits added answers the first question and is silent on the second R.

Outages are rarely total: a damaged cable is repaired in weeks or months, and the binding constraint is often the repair (permits for territorial waters, the availability of cable ships, the safety of crews in contested waters), which turns the problem into one of dynamics (Chapter 7) rather than topology. Capacity is not fungible: traffic between particular cities, with latency requirements, cannot simply move to a route around Africa. Terrestrial routes have chokepoints of their own, such as borders, power and permits, that the model's single letters hide. Failures are neither independent nor fixed: the regional shock in the model is one stylised correlation, and real risks change with events. And the model has no owners: who controls a landing station, a crossing permit or a repair ship is exactly what decides the real problem, and it is absent here.