urllib.py uses "*_proxy" variables to identify proxies
to use in the "getproxies()" function. However, it does
not use the corresponding "no_proxy" variable in the
"proxy_bypass()" method.

This affects any platform that can use "*_proxy"
environment variables to specify proxies. This includes
win32, which will use environment variables before the
registry if they are defined.

On win32 this causes the confusing behaviour of using
the registry for "proxy_bypass()" checking when
"*_proxy" environment variables are used to identify
the proxies.

Attached is a patch against python-2.3.2 that adds
support for "no_proxy". Note that win32 will correctly
favor using "no_proxy" over using the registry if any
"*_proxy" variables are defined.

Logged In: YES 
user_id=10273

Whups... failed to attach patch.

Logged In: YES 
user_id=10773

What is the state of this bug? Is something wrong with the
patch that it was not applied yet?

It is something annoying to unset $http_proxy manually for
some programs.

Logged In: YES 
user_id=10273

This bug is still present in current versions of Python.

The biggest problem with this patch is it only addresses urllib, 
when urllib2 has the same problem.

I have been meaning to put together a new patch that fixes 
both urllib and urllib2, but haven't had time or external 
motivation yet.

So, it is two years later now, will this ever be fixed? Is there a way to disable to use of proxies completely without having to change the enviroment when http_proxy is set? It is a bad user experience, when one is only able to change a proxy but not to disable the use of proxies completely.

Added docs and committed as r60134. Thanks for your patch!