Suppose a customer wants to import fifty reading lists from another application. Sending fifty requests adds coordination and network overhead, but putting all fifty writes in one request creates new questions. If the seventeenth item is invalid, should the other lists exist? If the client never receives the response, which items can the client safely submit again?
Bulk and batch APIs group work while preserving understandable outcomes.
This chapter covers request boundaries, atomicity, item results, authorization, retries, and asynchronous processing using a fictional bookstore API over HTTPS.