While GUI programs on Linux usually support Autoconfiguration proxy servers, most tools do not. Notably, cURL doesn't perform well. The WPAD protocol gives admins the ability to specify whether websites are to be accessed on a granular level. They can deterime the correct proxy server or direct connection based on the user's IP address as well as the destination domain name. Thus, when outside of the company network, this script will be able to tell that the user is connected directly to the internet.

However, the http_proxy environment variables are all-or-nothing. Either all of your traffic is routed through the proxy (if it's set) or no traffic is (if it isn't set). Furthermore, due to the nature of environment variables, it isn't possible to change this variable once the process has started. This makes VPN use particularly frustrating, as you constantly need to juggle whether you're on the network on not.

The proxyd project aims to alleviate this issue. It is a thin proxy server that bridges your programs to servers by utilising the libproxy library. libproxy is meant to be a smart proxy resolver capable of reading your system's settings and figuring out which proxy you need, if any. Any HTTP proxy compatible program that uses http_proxy will work with proxyd and get smart.

At work / VPN:
curl -> proxyd -> work proxy -> internet

At home / coffee shoppe:
curl -> proxyd -> internet

proxyd is at the moment a Python 2.7 program developed for RHEL 7. This is likely to change in the near future. To get it to work with GNOME 3, you need to install:

# yum install -y libproxy-{gnome,kde,networkmanager,python}

This gives you the necessary libproxy modules so that it can detect your chosen DE's settings, as well as the necessary Python 2.7 bindings.

Then, you can launch the server on port 1234 (likely to change):

./main.py

To test it:

./test.sh curl -L google.com

• At the moment, the CONNECT proxy command isn't supported. It is required for HTTPS connections.