dominant/TODO.md

Alpha

• Evict cache entries in order to provide bounded memory usage
  • LRU cache?
  • Hard limits
• Bound computational complexity
  • Context-aware machinegun?
  • Periodical history snapshotting?
  • Reuse cache entries where possible

Release

• Simplify cache interfaces
• Encode machine events and responses with a schema-aware protocol
  • Thrift / compact protocol?
• Fix potential race when a Repository request getting processed earlier than the start machine request being issued.
• Don't go to the cache for getting Head reference