Okay, there's a test case of web.py:

Server codes are following:
import web
class index(object):
    def GET(self):
        yield 'hello\n'
        yield 'world\n'
        time.sleep(60)

client is Python interpreter
>>> resp = urllib.urlopen(URL)
>>> resp.readline() # will be 'hello'
>>> resp.readline() # will be 'world'
>>> resp.readline() # huh, it's blocked, and we to agree with it.

>>> # but to use urllib2 will another behavor.
>>> urllib2.urlopen(URL).readline() # huh, it's blocked even if 'hello' and 'world' returned yet. Because urllib2 uses a 8KiB buffer on socket._fileobjece within urllib2.py, it read 8K data to buffer first.