Experimental Evaluation¶
crushr architecture was selected through destructive testing.
Method¶
- generate archive
- apply deterministic corruption
- measure recovery
Metrics¶
| Metric | Meaning |
|---|---|
| recovery_rate | % of data recovered |
| name_retention | % of filenames preserved |
| false_positive | incorrect recovery |
Result summary¶
- distributed identity consistently outperformed central metadata
- mirrored dictionaries retained names under partial corruption
- FORMAT-15 failed due to leadership dependency
Conclusion¶
Architecture is evidence-driven, not theoretical.