SkeinDB
A single-binary database engine written in Rust. SkeinDB combines cell-interned MVCC storage, MySQL compatibility, partial PostgreSQL support, an embedded administration surface, and 20 research tracks in one executable.
One runtime for implementation and evaluation
The project keeps storage, wire protocols, typed RPC methods, research features, and administration in one process. This makes system behaviour easier to inspect than an equivalent stack distributed across services and sidecars.
The storage design interns identical cell values across rows and versions. The engine exposes MySQL compatibility, an early PostgreSQL v3 baseline, HTTP RPC, and QUIC transport.
- 01Cell-interned MVCC and segment-backed row storage
- 02MySQL corpus-backed compatibility and partial PostgreSQL support
- 0320 opt-in research tracks with dedicated administration surfaces
- 04Single executable with no required service cluster
Research questions embedded in a working engine
The research agenda covers learned indexes, differential privacy, forensic audit, CRDT merge functions, incremental views, QUIC-native protocols, vector retrieval, performance replay, Wasm query operators, and energy-aware compaction.
A systems paper describes the architecture, cluster control plane, research tracks, and web-native administration stack.
Evidence boundary
SkeinDB is an active pre-1.0 system. Implemented and tested features are distinguished from prototype tracks; compatibility is described as partial where full parity has not been established.