In general, fork is useful when a saga needs to start a non-blocking task. Non-blocking here means: the caller starts the task and continues executing without waiting for it to complete.
There is a variety of situations where this can be useful, but the 2 main ones are:
- grouping sagas by logical domain
- keeping a reference to a task in order to be able to cancel/join it
Your top-level saga can be an example of the first use-case. You’ll likely have something like:
yield fork(authSaga);
yield fork(myDomainSpecificSaga);
// you could use here something like yield [];
// but it wouldn't make any difference here
Where authSaga will likely include things like:
yield takeEvery(USER_REQUESTED_LOGIN, authenticateUser);
yield takeEvery(USER_REQUESTED_LOGOUT, logoutUser);
You can see that this example is equivalent to what you suggested, calling with fork a saga yielding a takeEvery call. But in practice, you only need to do this for code organisation purposes. takeEvery is itself a forked task, so in most cases, this would be uselessly redundant.
An example of the second use-case would be something like:
yield take(USER_WAS_AUTHENTICATED);
const task = yield fork(monitorUserProfileUpdates);
yield take(USER_SIGNED_OUT);
yield cancel(task);
You can see in this example that the monitorUserProfileUpdates will execute while the caller saga resumes, and gets to wait to the USER_SIGNED_OUT action to be dispatched. It can in addition keep a reference to it in order to cancel it when needed.
For the sake of completeness, there is another way to start non-blocking calls: spawn. fork and spawn differ in how errors and cancellations bubble from child to parent saga.