* Fix quadratic performance in number of response .read() calls
 * Add a performance test (not run from test.py, will run from release script)

4 files changed