Sorry, hit submit by accident :) Flow control is hop by hop, not end to end, and so all we care about is detecting the hop-by-hop flow control issues. Specifically, we care about detecting when the remote side could not send because its flow control window was zero. -=R On Wed, Apr 16, 2014 at 9:50 AM, Roberto Peon <grmocg@gmail.com> wrote: > Flow control is hop by hop, not end to end. > On Apr 16, 2014 8:39 AM, "David Krauss" <potswa@gmail.com> wrote: > >> >> On 2014–04–16, at 11:07 PM, Johnny Graettinger <jgraettinger@chromium.org> >> wrote: >> >> I think there's a misunderstanding here. Priorities are about the >> relative ordering of multiple *ready-to-write* streams. A sender can and >> certainly should send a dependent stream frame, if that's what's available. >> >> >> I didn’t say anything about priorities. A sender should only send a >> dependent stream frame if all its dependencies are blocked. Otherwise, the >> dependent stream is blocked. >> >> Can you give an example of when a unidirectional BLOCKED frame would be >> sent? >> >> An empty DATA frame will simply be coalesced (ignored) by the first >>> intermediary. >>> >> >> That's fine. Flow control is also hop-by-hop. >> >> >> It’s not fine because such a debugging/tuning facility stops working if >> any ISP, load balancer, or firewall employs a proxy. >> >> What is it you intend to accomplish? There may be a misunderstanding. >> >>
Received on Wednesday, 16 April 2014 17:02:50 UTC