gcloud.iterator.Iterator is a base class, relying on a path for the API method. I think it would be clearer if it wrapped around a method of a client (or other domain object), where that method takes page_token, page_size, etc., along with other named arguments in **kw. We could then ditch all the derived classes, and just use the one, common implementation. E.g.:

from gcloud.iterator import Iterator
from gcloud.pubsub import Client

client = Client()
topics = list(Iterator(client.list_topics))
subs = list(Iterator(client.list_subscriptions, topic_name='my-topic'))