Cancellability: Why Should a Started Process Still Be Stoppable?

Cancellability is the ability of a process to accept a stop request before it finishes naturally and settle into a clear, safe state. It differs from reversibility. Reversibility asks whether a completed result can be undone; cancellability asks whether unfinished work can cease.

Imagine a batch job that will generate one hundred files. The user discovers after the fifth file that the parameters are wrong. A real cancellation cannot merely close the progress window. It must prevent new work from entering the queue, let the smallest active step stop at a safe boundary, and identify which of the first five files are complete or incomplete. Otherwise the interface appears quiet while the background continues consuming time and writing bad results.

Cancellation does not always mean instantaneous termination. If a step must remain atomic, the system may finish or roll back that small unit before stopping at an intelligible boundary. The concept matters because starting should not automatically become an inescapable commitment. The longer and more costly a process is, and the more it changes external state, the more its stopping behaviour must be part of the design rather than a patch added after failure.