Re: Working Group Last Call: Bootstrapping WebSockets with HTTP/2

On April 4, 2018 1:17:10 AM GMT+08:00, Alessandro Ghedini <alessandro@ghedini.me> wrote:
>On Thu, Mar 29, 2018 at 08:48:28AM +1100, Mark Nottingham wrote:
>> Hi everyone,
>> 
>> Patrick (as editor) has incorporated the discussion from London and
>believes this is ready for WGLC; there are no open issues.
>> 
>> Please have a look at:
>>   https://tools.ietf.org/html/draft-ietf-httpbis-h2-websockets-01
>> 
>> ... and bring up any issues on-list or on its issues list; likewise
>statements of support (or otherwise) for publication on-list would be
>appreciated.
>
>Hello,
>
>Need a clarification regarding section 5. It says:
>
>   Implementations using this extended CONNECT to bootstrap WebSockets
>   do not do the processing of the [RFC6455] Sec-WebSocket-Key and Sec-
>  WebSocket-Accept headers as that functionality has been superseded by
>   the :protocol pseudo-header.
>
>does that mean that a transparent proxy (say, a CDN, as opposed to a
>traditional HTTP CONNECT proxy) that proxies WS connections and speaks
>HTTP/2
>with the client but HTTP/1 with the origin server, is expected to
>generate
>-Key and validate -Accept itself? Might be worth mentioning this in
>section 7
>as well.

Yes it means that's solely a problem for the h1-speaking side to fake up.  The h2-speaking side doesn't have that ws upgrade handshake concept and no state from it is used on that side.

-Andy

>Cheers