Data format

This page will define the on-disk binary contract for Lakestream storage: the WAL object format used for durable, low-latency appends, and the compacted object format used for long-term storage in row or Parquet layouts. It matters because interoperability starts here — any implementation that reads and writes these two object shapes correctly can participate in the standard, regardless of which engine wrote them.

Planned coverage

• The WAL object binary format
• The compacted object binary format
• Row layouts vs Parquet layouts, and when each applies

Source material

Documented in the ursa-storage repo, at project-docs/data-storage-spec.md.