This is establishing the API for a future optimization. We want to
be able to offer true socket-level caching, but can't do that with
httplib2 right now. For now, we will just fake the optimization
by returning an iterator over the image body, which happens to already
be fully loaded into a string.
Change-Id: I2d36e3cdd45b26d7c7c27ba050bf6a4b5765df6c
Rather than depend on magic, I would prefer that we explicitly call
two different request methods: json_request and raw_request. The
former will encode/decode request bodies to and from JSON, while
the latter will not.
Change-Id: I6a429a5975993f71df85df55f11c5d51c050c289
Brian Waldon