QUIC is certainly interesting, but I can not agree with the headline. IMHO HTTP over QUIC needs a very different implementation than HTTP/2 over TCP/SSL. The ma

QUIC is certainly interesting, but I can not agree with the headline. IMHO HTTP over QUIC needs a very different implementation than HTTP/2 over TCP/SSL. The majority of frame types and frame handling mechanisms are no longer needed and even if they are, they will be used in other ways. That's similar to the situation that HTTP/2 is not simply HTTP/1.1 over a multiplexing layer that gets introduced on top of TCP but also a completely different thing, as the multiplexing layer and the HTTP related mechanisms (header transmission, compression, etc.) are tightly interacting there. My current estimation is that webservers would need to have 3 completely different implementations of the HTTP stack (for HTTP/1.1, HTTP/2 and HTTP/QUIC), where only the most upper layer (interfaces for sending requests and respones) can be shared, the remaining stuff will be very different.