- From: Ilari Liusvaara <ilari.liusvaara@elisanet.fi>
- Date: Mon, 2 Sep 2013 21:18:03 +0300
- To: Adrien de Croy <adrien@qbik.com>
- Cc: William Chan (ιζΊζ) <willchan@chromium.org>, Martin Thomson <martin.thomson@gmail.com>, James M Snell <jasnell@gmail.com>, Ryan Hamilton <rch@google.com>, HTTP Working Group <ietf-http-wg@w3.org>
On Sun, Sep 01, 2013 at 09:37:38PM +0000, Adrien de Croy wrote: > > in all this didcussion about CONNECT, I presume that references to > server, are to be interpreted as proxy? > > Since CONNECT is for proxies only. > > Or is it being proposed this could be useful for end servers? Copying the diagram: Client <-- HTTP2 stream ---> Proxy <-- TCP connection --> Server The semantics were written from the point of view of the proxy. But I think similar semantics could be applied to upgrade, where: Client <-- HTTP2 stream --> Server <-- unspecified --> Service The transport between server and service is unspecified because it is internal implementation detail not visible to the client. Of course, that unspecified transport would have to internally define concepts of end-of-data and reset. But if it can run from HTTP/1.1, it must anyway. -Ilari
Received on Monday, 2 September 2013 18:18:28 UTC