httplib: response.read() vs response.read(size)

[Doug Fort]

We're developing a website load testing application
(http://www.stressmy.com).  We give httplib a pretty good workout.  And
generally get excellent results.  Recently, we ran into a pair of unrelated
sites that use some form of graphics server (Netscape-Enterprise/4.1).
There, our simulated users hung up waiting for a read on the socket in the
httplib.HTTPResponse.read().  The code looks like this (with a lot of
deletions)

self._client = httplib.HTTPConnection("", None)
...
response = self._client.getresponse()
...
self._data = response.read()

I was able to work around this with the following:

if response.length:
    self._data = response.read(response.length)
else:
    self._data = response.read()

My question is:  should this be moved up into httplib.HTTPResponse.read()?
I'm posting this here, instead of sending a bug or a diff to SourceForge
because I really don't understand the new HTTP 1.1 state machine in httplib,
and I don't want to aggravate the developers with bogus bugs or dumb patches
while they're working on 2.1.
--
Doug Fort
Senior Meat Manager
Downright Software LLC
http://www.dougfort.net