Base
- Lookup software related attack_ids in the revoke map
- Updated Top level domains list
- Use document versions to get and save file atomically instead of using locks
Core
- Properly handle invalid submission classification
- Use document versions to get and save file atomically instead of using locks
- Perform archiving in chunks to prevent timeouts
- Pre-compute expiry daterange instead of using a datemath inside the expiry lucene query
Frontend
- Do not auto-retry on quota exceptions
- Auto-retry API calls when API server is down
- Make auto-retries use an exponential backoff
Sevice Base
- Stop polluting logs from empty extracted/supplementary files
UI - API Server
- Use one global filestore per process
- Re-use globally loaded config, filestore and classification engine
- Improve ingest API speed by removing extra filestore calls and by optimizing the sha256 submit usecase