Delivery
Publishers write state over HTTP. A delivery is the routing layer that decides which matching object updates reach subscribers. Clients connect to a delivery channel; they receive the streams selected by that delivery’s object patterns.
HTTP remains the source of truth. Delivery notifies you that something changed—you can always re-fetch the latest or a specific version when you need the authoritative value.
How it fits together#
Create delivery (object patterns) → clients connect → publisher sets state → matching updates arrive
- Configure a delivery once from a trusted environment (API: Delivery).
- Connect subscribers with credentials allowed by policy.
- Publish state as usual; only objects that match the delivery’s patterns are pushed.
Use separate deliveries when different clients need different feeds—for example a dashboard on a chart namespace versus a narrower mobile feed—so each connection only sees what it should.
Channels#
BigState delivery configurations support several channel types. This section documents how to subscribe over WebSocket (type 1), which is what the getting-started guides and tutorials use.
For creating and managing delivery resources (name, type, object patterns), see the Delivery API reference.
Before you connect#
- An existing delivery id and object patterns that cover the streams you care about
- Credentials scoped for the subscriber (prefer a dedicated principal / key—not a privileged publisher key in the browser)
- Optionally, a latest-state HTTP read on startup, then live updates over the socket
A first end-to-end path is in Deliver and read state. Hands-on apps are under Tutorial.