On Tue, Apr 22, 2014 at 6:42 PM, Mark Nottingham <mnot@mnot.net> wrote: > > On 23 Apr 2014, at 11:39 am, Ryan Hamilton <rch@google.com> wrote: > > > > > > > > > > > On Tue, Apr 22, 2014 at 6:12 PM, Mark Nottingham <mnot@mnot.net> wrote: > > > > On 22 Apr 2014, at 4:12 am, Martin Thomson <martin.thomson@gmail.com> > wrote: > > > > > On 19 April 2014 23:36, Amos Jeffries <squid3@treenet.co.nz> wrote: > > >> The draft wording however is not limited to "proxies". Which was my > > >> initial report of there being a problem. > > > > > > Mark already suggested that the wording needed to be changed from > > > "intermediary" to "proxy". I think that suffices. > > > > It might also be helpful to note that a client with a configured proxy > isn't expected to use alternative services (no matter how discovered). > > > > That doesn't sound right to me. Couldn't the client use a CONNECT > tunnel to reach the alternate service? If the alternate service is HTTP/2, > and the main service is HTTP/1, then there could be a big performance win > from using the alternate service through the proxy, wouldn't there? > > ... isn't expected to use alternative services (no matter how discovered) > to establish new connections; however, existing connections through the > proxy might be affected (e.g., CONNECT tunnels). > I'm not following, which suggests that I'm missing something. :> Imagine that a client is visiting http://server/ via the proxy, which serves up an alt-service header suggesting server:443. The client would then initiate a new CONNECT tunnel through the proxy to server:443. Was that what you meant? > ;) Cheers, Ryan
Received on Wednesday, 23 April 2014 03:00:57 UTC