This is a very clear and well written document. I believe it is ready to be published. I spotted only one minor editorial issue in the security considerations: [...] connections that use DATAGRAM frames might be distinguished from other frames using the different response to loss. The phrase "connections might be distinguished from other frames" feels a bit odd. How about this instead: [...] connections that use DATAGRAM frames might be distinguished from other connections due to their different response to frame losses. /js