Expand description
Where a computed plan waits for its approval.
A plan carries every byte it will write, and some of those bytes come out of the target. So the store is owner-only, it lives under the state root rather than in the repository, and a plan is ephemeral: it is never committed, and never pasted into a forge.
Identity is the fingerprint and nothing else. Identical inputs while an executable plan exists reuse it, so an operator who plans twice approves one thing. A terminal result moves the plan out and frees the fingerprint, so identical inputs after a success plan again into a fresh directory rather than meeting a stripped one.
The lifecycle is fixed rather than configurable. An unapplied plan expires after seven days. A terminal result is kept for thirty. A run that needs recovery keeps everything recovery needs, whatever the calendar says.
Structs§
- Operation
Outcome - What one operation did.
- Plan
Directory - Where one plan’s own directory sits.
- Postcondition
Outcome - What one postcondition found.
- Result
- What one apply did.
- Store
- The owner-only store under the state root.
Enums§
- Disposition
- How one apply ended.
Constants§
- PLAN_
TTL_ DAYS - How long an unapplied plan waits before it expires.
- RESULT_
SCHEMA - The machine schema a result declares.
- RESULT_
TTL_ DAYS - How long a terminal result is kept.
Functions§
- redact
- The plan as a result keeps it: no absolute target path.