Three Clocks Pick Local or Server
A model call should stay local until three clocks agree. Latency, secrets, and offline need are those clocks. A free server wins only inside that narrow overlap. The laptop is a workshop with the lights already on. The network is a ferry with a posted schedule. You board when the cargo is clean and the crossing fits the deadline. Habit is a bad dispatcher for this choice. A fast demo last week…
Three distinct clocks determine whether a task should run locally or on a server. The first clock is latency, which measures how quickly a task can complete. The second clock is the secret clock, which ensures sensitive data is not exposed to untrusted servers. The third clock is the offline clock, which handles tasks that must complete even when there is no network connectivity.
A free server can handle only a limited set of tasks that meet specific criteria, such as having no secrets and being able to finish within the allocated time budget. If the measured round trip time exceeds this budget, the task cannot be completed on the server and must run locally. If the task contains sensitive data or requires offline execution, it should always run locally, regardless of the round trip time.
A local runner should be given the same time budget as the network, and a timeout indicates that the local execution could not meet the deadline. The policy module proposed in this article provides a framework for deciding whether to run a task locally or on a server based on these three clocks.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
