I am confused about the num_retries differences between the old googleapiclient and the new google-cloud-python.

I am able to specify num_retries for all API requests with the old googleapiclient by either

1. specifying num_retries for all requests
2. or supplying own requestBuilder for the googleapiclient.discovery.build which subclasses the original googleapiclient.http.HttpRequest and sets num_retries for all API requests.

The new google-cloud-python allows num_retries just for the storage blob upload. Is there other way to specify num_retries for other API calls? I think that all API calls (including Datastore, etc.) may fail by HTTP 500 or 429 and then they need to retry the requests, e.g. see https://cloud.google.com/storage/docs/exponential-backoff

Also, the googleapiclient.http._should_retry_response() seems different than google.cloud.streaming.http_wrapper.make_api_request(). E.g. the old one examines the response JSON payload, too.