Rubric: L-RELIABILITY
Lens: L-RELIABILITY · Density: D-MED · Axes: REL, ROB, RCV
Criteria
- Error handling completeness on I/O and RPC boundaries.
- Idempotency / retry / timeout behavior.
- Data integrity under crash (write ordering, fsync, transactions).
- Degradation vs silent corruption.
- Fail-closed vs fail-open at safety boundaries.
Citations
| Work | Point |
|---|---|
| Google SRE Book | Embracing risk; error budgets (conceptual) |
| Kleppmann — DDIA | Reliability; disk/fsync; replication truths |
| ISO/IEC 25010 | Reliability characteristic |
Pros / cons
| Pros | Cons |
|---|---|
| Distinguishes “works on happy path” from production | Requires domain reading; tests alone insufficient |
| Crash consistency is often missing in reviews | Over-indexing on distributed systems lore for single-process tools |
Recommendation bias
Prefer evidence of corruption refusal and explicit degradation over “best effort continue.”