Remote / Server-derived state

When not to use Remote

  • Local-only state. Use the Model and update; there is no server truth to cache.
  • Offline writes and convergence. Use foldkit-sync; it owns client-authored operations and orders them through a durable log. When the rows edited are too many to replicate, Remote still reads them and the edits are laid over them: Editing server data through a journal.
  • Request-level caching of one expensive endpoint. Effect's PersistedCache fits (Request → Result); Remote is Entity + Field → Value and normalizes across requests.
  • Peer-to-peer or CRDT replication. Out of scope.
  • A general-purpose database. Remote holds an application-facing cache; it is not a query engine. foldkit-remote-drizzle compiles the selection and query it already understands into SQL, and no further.