End-of-week reading: The classic end-to-end principle paper that guides the interface design of the Internet. It's a great paper. The principle can also be understood by asking two simple questions: a) Does a lower-level implementation sufficiently solve the problem? If not, don't. That corresponds to the end-to-end argument. b) Is the lower-level implementation necessary to unlock performance (e.g., the early-retry protocol in transmission) that wouldn't be possible otherwise? If no, don't.