Please don't call it WebSocket. It just confuses people. I used the name "WebUDP" for this last year. Although it never really caught on, so maybe another name is needed.

Let me own this bug for a while.

@ricea I think you had written an IETF draft as well sometime last year when we (Citrix) requested this feature. Hopefully WebUDP or equivalent gains some traction!

Hi santosh.sampath@, have you looked at https://github.com/w3c/webrtc-quic? Will it solve your problem? Do you want https://github.com/w3c/webrtc-quic/issues/26?

background docs: https://docs.google.com/document/d/1RUnws_QT_ZVJqiPJ8XT1tVg3HH921BJ3-BXVBzYlCjA/edit#heading=h.8rxi7pjhgqip
https://drive.google.com/u/0/open?id=1E7Z5VW3A9rHQSz0KEi1Imsv1HFXfRaXCoq_n6n2GAuc

brad, can you comment on what the requirement for this API?

I think the base requirement is a web API for unreliable, unordered packet delivery. Practically, I expect the best approach is a thin layer over the proposed QUIC DATAGRAM frame, ideally with the ability to reuse existing QUIC connections, expose ack information and enforce/ensure single packet messages.

sam, can you help investigate and provide an estimation on the work.