Coding classes in County Durham

The learner builds a toy line of six single-track sections, with passing places at each end and one in the middle where trains can wait side by side. Each section can hold only one train. Trains set off from either end at random times. The first rule anyone tries is the obvious one: move into the next section whenever it is empty. Run it, and it often works. Run it a thousand times with different random timetables, and a pattern appears. Sooner or later an eastbound train sits in one section wanting the next, while a westbound train sits in that next section wanting the first. Both wait for ever.

The fix is the token rule, an idea railways have long used on single lines: a train may only leave a passing place once it holds the right to every section up to the next passing place, all at once. If any of them is taken, it waits where there is room to wait. With that one change, not a single run out of five thousand deadlocks. The price shows in the last column: trains sometimes wait at a station when they could have crept forward, so with eight trains the line takes 21.7 steps on average against 16.2 for the few naive runs that happened to finish.

This is exactly the deadlock that computer scientists study. Four things together cause it: something only one user can hold at a time, users holding one thing while waiting for another, no way to take a thing away from its holder, and a circle of waiting. Break any one and deadlock is impossible. The token rule breaks the second: nobody holds part of what they need while waiting for the rest. Databases, operating systems and online games all use versions of the same trick.

So a County Durham teenager should learn to code in 2026, two hundred years on from Locomotion No.1, to be the one who spots the jam before the trains do. The longer argument is in why teenagers should still learn to code in 2026.