Regarding replicators; are they intended to directly replicate the state from one service to another. Like syncing the JSON structure directly into local storage or the file system, or are replicators fit for any situation where data has to be retrieved or sent to a third-party? E.g. via some SDK that maybe expects a different format for its input data than that of the store’s state data. I.e. when there’s a need to “translate” some prop change to an HTTP request and vice-versa.
Or should this rather be done as traditionally, by having the action creators call the SDK directly and dispatch other actions upon fulfilment/failure of the async request (via middleware like “thunk”)?
It would be nice if the first is true, since that would make testing a breeze and allow for consumers of some provider to drop in a custom version of said SDK on-demand etc.