Yes, the use-case is where if we don't have current data, the client registers a request and then we later serve that request. Essentially, a server deferred reply to the client.

Thanks! That was useful.

We implemented something very simple when we had to do public async type API calls, user does transaction, gets a response back saying we'll ping him later since we don't have the data yet.

Once we have the data, we do a POST request to the client with just a transaction ID, we expect the client to respond with an OK if they received our POST request, once we receive the OK from the client, we stop trying to publish to the client (if we don't get the OK after multiple attempts over two days, we send an automated email saying their endpoint is down and that we're trying to notify them that transaction id xyz is ready for querying)

Now once the client sent us the OK, we have the data available and the client simply needs to do a query request using the transaction id they've received.

This is a public API that we are implementing.