A buoy is visited, just slowly and at a cost

Mru Field is the same core as the flight software, for systems on Earth: ocean observatories, buoys and moorings, polar and seismic stations. The repository holds a specification dated 21 October, a README, a license, and nothing that runs. Status: planned, not built. Code lands after the second OPS-SAT package and before the first deployment, a twelve-week ocean test planned for 2027 on a spar buoy off Genoa.

The spec has one design decision and says so: inputs are read once. The supervisor acquires one input record per cycle and hands the same bytes to every replica, so replicas cannot disagree because they sampled a sensor at different moments. It runs under systemd and does not replace it. systemd restarts a process that died. The supervisor catches one that is alive and wrong, alive and hung, or restarted from zero after a reset.

The first word of the README was “unattended”. The buoy gets one to three boat trips a year, a day trip from Genoa on a ten meter boat, only in flat sea. It is visited. The word is now “hard-to-reach”, defined in the spec as a system where a repair visit is slow, costly or impossible, one that may be visited once a year or never. The flight README, the profile card and the website changed the same morning.

Unattended described the software. Hard-to-reach describes the problem.