@handrews very diligently improved the rules and explanations around percent-encoding in Appendices D and E of the 3.2 spec. Should these improvements be downported to 3.1 (were appropriate)?