> Antoine, absolument.  Please see attached file bug.py

Ah, thanks, I see. But this is not caused by this issue.
It seems to come from 12442ac3f7dd.

> Instead we should simply define it for python, and in accordance to
> established tradition, namely that defaulttimeout prevails.

That would be the second behaviour change in two versions. I think we
could/should simply fix the advertised behaviour, as in
http://docs.python.org/dev/library/socket.html#timeouts-and-the-accept-method

> Btw, defaulttimeout(None) doesn't mean that that there is no default,
> it means that the default is "blocking."

Sure, but generally it means that the user didn't *set* a default at
all.