Reach a workspace replica on an authorized Host, including your own laptop or phone.
An application's knowledge belongs to a workspace, even when its serving machines change. Routing resolves that workspace identity to a Host serving an eligible workspace replica, then resolves the peer's connection address.
Callers can use the workspace reference instead of maintaining a separate machine URL for every participant in a distributed Use Case.
A Host can run on your own laptop, phone or another mobile device. Keep your most precious workspaces' serving replicas on hardware you control, and authorize only those Hosts to decrypt and query their contents.
Peer connectivity supports NAT traversal and encrypted relay paths. This helps a personal Host participate behind a home router or restrictive network without requiring a conventional publicly reachable server endpoint.
Direct paths avoid a relay hop when available. Relayed paths provide another connectivity option, with their own latency, capacity and infrastructure dependencies. Network policy still determines which paths can work.
Multiple eligible serving Hosts provide more placement choices. A reachable custodian holding encrypted blocks is a storage participant; executing a Story requires an authorized Host with the relevant state, keys and capacity.
Other workspaces used by the same application can have serving replicas elsewhere. Your private workspace's encrypted blocks can also live with independent custodians for durability, while plaintext processing stays on your personal hardware.
If only your phone may decrypt a workspace, routing preserves that decision when the phone sleeps. A storage copy cannot silently acquire query authority.
High availability covers serving redundancy. Native Story calls describes execution over the peer network.
Developer reference
Workspace hosting lookup filters by requested workspace and reachability. The selected record names a node identity; node-address resolution supplies connection information. Reject a record for a different workspace rather than routing on a coincidentally reachable address.
The lookup identifies a candidate. Admission still checks workspace authority and the requested operation. Reachability observations can become stale between discovery and connection; they are not a service reservation or proof of replica completeness.
Transport can establish direct or end-to-end encrypted relayed connectivity. Relays forward protected traffic without receiving workspace decryption authority. Endpoint reachability, serving eligibility and operation completion are three separate states.
Changing Hosts must respect branch/commit requirements and key-delivery policy. A failed connection is evidence about a path; a timed-out mutation needs an execution-outcome check before any retry. Replica count alone supplies neither automatic replay safety nor an exactly-once guarantee.