Comment 23 for bug 392195

Olivier, sorry I didn't respond any sooner I had only seen the comment on Launchpad. I certainly can make the change. I had opted to use a separate test because it does cause one to wait substancially less for the standard connection to fail (since you can wait a long time if a firewall just drops packets rather than sending an ICMP reply). I however know that the test itself was kind of poorly done, with an arbitrary timeout...

Thanks for the response about using the HTTP transport by default for proxy connections, it does make sense. That said, I'd then also remove "proxies" from the standard connection since it would always be empty in that case.

/ Matt