Agent Architecture and Orchestration
Shared State Management — References
Shared state management › References
The concurrency and consistency literature this problem is a special case of, plus the mechanisms used to make shared records recoverable.
Consistency and concurrency
- Jepsen: consistency models — a precise map of what guarantees stores actually offer, and the vocabulary for stating what participants are promised about each other's writes.
- Linearizability: A Correctness Condition for Concurrent Objects (Herlihy and Wing, 1990) — the definition that claim-and-expire mechanisms are implicitly reaching for. Cited by name.
- Time, Clocks, and the Ordering of Events in a Distributed System (Leslie Lamport, 1978) — why wall-clock timestamps cannot order concurrent writes, which matters as soon as provenance includes a time. Cited by name.
- Designing Data-Intensive Applications (Martin Kleppmann) — the practical treatment of lost updates, write skew, and the isolation levels that do and do not prevent them. Cited by name.
Recording rather than overwriting
- Event Sourcing — persisting the sequence of changes as the system of record and deriving current state from it, which is the append-and-project arrangement in its original form.