Client-side ID generation (e.g. UUIDs) can be very useful. For one, it lets the client send an entire batch payload of one or more inserts, plus updates that reference the new inserts, in a single round trip. It doesn't need to wait to get the inserted IDs back to make updates, or send more requests to other services. All it needs is an OK back. This is very convenient in systems where, say, you ingest a bunch of data that needs to be dumped into a final database. The entire batch can be ready, including IDs, at the sender side.
This is particularly useful in some other scenarios such as distributed, eventually-consistent systems that are able to resolve concurrency conflicts. Send out changes optimistically (with those pregenerated IDs), resolve conflicts by untangling the ones that didn't work.
This is particularly useful in some other scenarios such as distributed, eventually-consistent systems that are able to resolve concurrency conflicts. Send out changes optimistically (with those pregenerated IDs), resolve conflicts by untangling the ones that didn't work.