Persistence layers in SpaceManager

Purpose

SpaceManager loads and saves a space through one or more persistence layers. SpaceRunner owns the runtime wiring per space (load, merge, sync, save). A layer stores RepTree operations and optional secrets.

How layers are chosen

The client creates layers from the workspace URI. Examples:

• local:// uses IndexedDB only.
• file:///path uses IndexedDB + filesystem.
• https:// uses IndexedDB + RemoteSpacePersistenceLayer (socket.io sync).

See createPersistenceLayersForURI for the mapping.

How loading works

SpaceRunner connects all layers in parallel. It builds a space from the first layer that returns operations. It then merges operations from other layers that succeed.

This keeps startup fast while still converging state. If all layers fail, the space load fails.

How saving works

SpaceRunner writes new operations to every active layer. If a layer stops working, the remaining layers still persist data.

Implications

• Use more than one layer for redundancy.
• A fast layer makes startup faster.
• A slow or failing layer should not block the workspace.

Adding a new layer

Implement the PersistenceLayer interface. Register it in the URI mapping.

Remote sync is a layer that can:

• Fetch ops on load (state-vector diff).
• Push local ops to the server.
• Listen for remote ops and merge them.

Keep IndexedDB for offline caching.