Barry A. Warsaw added the comment:
On Jul 10, 2013, at 05:53 AM, R. David Murray wrote:
>Please see issue 14425 and then let us know if this is still valid. It's
>been a while since I looked at the code and I no longer remember the details,
>but I seemed confident in my conclusion at the time... :)
I'm not sure I completely understand the arguments in #14425 and #2451 which
it refers to. I could be wrong, but it seems to me that these are arguing
against using socket.getdefaulttimeout() in favor of the magic non-public
attribute that's currently being used. While I don't like that particular
aspect, it's not the real point of *this* bug.
Even if socket._GLOBAL_DEFAULT_TIMEOUT is the effective default for `timeout`,
I still think using that as the default value for the keyword argument is
wrong, both for usability of this API and for the documentation of it.
I don't see why `timeout=None` wouldn't be acceptable, but let's assume it
isn't. A documented, public sentinel would be acceptable and it could either
be defined as:
DEFAULT_TIMEOUT = object()
or
DEFAULT_TIMEOUT = socket._GLOBAL_DEFAULT_TIMEOUT
With the latter, the function's signature would then change to:
def urlopen(url, data=None, timeout=DEFAULT_TIMEOUT, ...)
and nothing functional would change. The documentation update would be much
more helpful too. If it weren't for the new module attribute, it could
probably even be backported.
Then users of this module could do something like this:
from urllib.request import DEFAULT_TIMEOUT, urlopen
def my_open(url, timeout=None):
my_timeout = DEFAULT_TIMEOUT if timeout is None else timeout
return urlopen(url, timeout=my_timeout)
----------
_______________________________________
Python tracker <rep...@bugs.python.org>
<http://bugs.python.org/issue18417>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
