ECSS Packet Utilisation — Suspending and Resuming (space-systems/ecss/e7041-suspending-and-resuming)
Use when the task is the hold and release of ECSS-E-ST-70-41C clause 6.18.4.6 -- telling a running on-board control procedure to stop where it is and keep everything it has, and letting it carry on later from exactly there.
Domain quick reference
- A hold is not a stop. Stopping ends the run and the procedure would have to begin again at its first step. Holding keeps the step pointer, the bound arguments and the engine slot, so a release carries on from where it paused. Every rule here protects that distinction.
- A hold lands on a step boundary, never inside a step. A step is the unit the procedure was written in, and halting half way through one leaves the spacecraft in a configuration the author never described: a valve driven and not confirmed, a heater commanded and not verified.
- So a hold commanded mid-step is pending, not effective. Between the command and the boundary the procedure is still running, and a ground that reads the acknowledgement as a hold is reasoning about a spacecraft that is still moving.
- The late hold. If the step in progress was the last one, the run ends before the hold can take effect. There is nothing left to release, and the honest report is terminated with the hold too late -- report it as held and the ground waits forever to resume something that finished.
- Holding is not idempotent. A hold commanded on a procedure that is already holding, or that already has one pending, is a refusal, not a no-op: the second command means the operator's model of the spacecraft is wrong and the refusal is what tells them.
- A hold costs an engine slot the whole time it lasts. It is not cheap parking. An unbounded hold is a slot lost quietly, so the longest hold the mission tolerates is a real limit and an over-run is a decision -- release it or abort it -- not a status.
- Hold time accumulates across holds. One procedure held three times for twenty minutes has cost an hour of its slot, and only the running total says so.
Workflow
- Validate the instance: a known state, a step pointer inside the procedure, and a hold time that is present exactly when the state says it is holding.
- Refuse a hold on anything that is not running, and refuse a second hold whether the first is landed or still pending. Name which.
- If a step is in progress, record the hold as pending and leave the state running. Report the pending state as pending.
- If no step is in progress, the hold lands immediately. Record the moment and the step it landed on.
- When a step completes, advance the pointer first, then decide. Past the last step the procedure has terminated, and a pending hold is reported as having arrived too late.
- Otherwise land a pending hold at the new boundary and record the step, so the release has somewhere to resume to.
- On release, require a held procedure, add the elapsed hold to the running total, count the release, and carry on from the held step. Refuse a release whose time predates the hold.
- Grade an open hold against the mission limit, treating a hold exactly at the limit as within it, and report the engine slot it is still holding.
- For an event sequence, refuse one that goes backwards in time, and keep each outcome with its moment and the state it produced.
Pitfalls
- Reading a hold acknowledgement as a landed hold. Mid-step it is pending, and the spacecraft is still executing while the ground thinks it stopped.
- Halting inside a step to make a hold prompt. The configuration left behind is one no procedure step describes, and no release path returns from it cleanly.
- Reporting a procedure that ran out as held. The ground then waits to resume a run that has already finished, and nothing on the link ever corrects it.
- Resuming from the first step. That is a restart wearing the word resume, and for anything already partly done it repeats actions that were never meant to run twice.
- Treating a repeated hold as harmless. The refusal is the only signal that the operator's model of the spacecraft has drifted.
- Assuming a hold releases the engine slot. The concurrency budget then reads free while it is not, and the next activation is refused by the engine rather than by the plan.
- Measuring only the current hold. The cost of a procedure held repeatedly is the accumulated total, not the last interval.
Behavior contract (gate 3)
The instance validation and its state and hold-time consistency checks, immediate versus pending holds, the pending hold landing at the next boundary, the late hold reported as termination, refusal of a repeated or pending-duplicate hold, refusal of a release on a running or pending instance, resumption at the held step, accumulated hold time and release count, engine slot retention while held, the mission hold limit with exact-limit acceptance, and the time-ordered event sequence are exercised by the gate 3 contract test: scripts/test_e7041_suspending_and_resuming.py against scripts/e7041_suspending_and_resuming_logic.py (stdlib unittest, offline). Run: python3 scripts/test_e7041_suspending_and_resuming.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
