2 Client and user ids

Sente uses two types of connection identifiers: client-ids and user-ids.

Client ids

A client-id is a unique identifier for one particular Sente client: i.e. one particular invocation of make-channel-socket-client!. This typically means one particular browser tab on one device.

By default, clients generate their own random (uuid) client-id. You can override this in your call to make-channel-socket-client!.

Note:

1. Each client chooses its own client-id with no input from the server.
2. By default, each browser tab has its own client-id.
3. By default, reloading a browser tab (or closing a tab + opening a new one) means a new client-id.

User ids

This is the more important concept in Sente, and is actually the only type of identifier supported by Sente's server>client push API.

A user-id is a unique application-level identifier associated with >=0 Sente clients (client-ids).

It is determined server-side as a configurable function of each connecting channel socket's Ring request, i.e. (fn user-id [ring-req]) => ?user-id.

Typically, you'll configure Sente's user-id to be something like your application's username: if Alice logs into your application with 6 different browser tabs over 3 different devices - she'll have 6 client-ids associated with her user-id. And when your server sends an event "to Alice", it'll be delivered to all 6 of her connected clients.

By default, Sente will use (fn user-id [ring-req] (get-in ring-req [:session :uid])) as your user-id function. You can override this in your call to make-channel-socket-server!.

Note:

1. One user-id may be associated with 0, 1, or many clients (client-ids).
2. By default (i.e. with the sessionized :uid value), user-ids are persistent and shared among multiple tabs in one browser as a result of the way browser sessions work.

Examples

Per-session persistent user-id

This is probably a good default choice.

1. :client-id: use default (random uuid)
2. :user-id-fn: use default, ensure that you sessionize a sensible :uid value on user login

Per-tab transient user-id

I.e. each tab has its own user-id, and reloading a tab means a new user-id.

1. :client-id: use default (random uuid)
2. :user-id-fn: use (fn [ring-req] (:client-id ring-req))

I.e. we don't use sessions for anything. User-ids are equal to client-ids, which are random per-tab uuids.

Per-tab transient user-id with session security

I.e. as above, but users must be signed in with a session.

1. :client-id: leave unchanged.
2. :user-id-fn: (fn [ring-req] (str (get-in ring-req [:session :base-user-id]) "/" (:client-id ring-req)))

I.e. sessions (+ some kind of login procedure) are used to determine a :base-user-id. That base user-id is then joined with each unique client-id. Each tab therefore retains its own user-id, but each user-id is dependent on a secure login procedure.