In my preliminary design report I proposed forking new copies of the proxy for each client connection, allowing the proxy to handle high loads. Upon closer inspection, however, I realized that a non-blocking event loop would serve me better, giving greater performance because of the extra overhead needed to create processes when forking and avoiding any synchronization difficulties. On the other hand, writing correct non-blocking code is quite complex, giving rise to new concerns.
The design section contains three subsection, describing the general proxy design, design of the request/response handling and cache design respectively.