recondb is an eventually-consistent multi-master key-value store. the core data structure is an ORSet.
recondb uses an experimental ORSet diffing mechanism to optimize propagating changes to known peers. new or stale peers must complete a full merge in order to maintain consistency.
- data is bencoded and compressed for storage
- records may have hierachical keys
- values are ordered based on their key
- data is stored in ipfs under an ipns key that is generated per-database
- how is this data queried? (graphql?)
- how does indexing work?
- should sharding be considered?