- From: Martin Thomson <martin.thomson@gmail.com>
- Date: Mon, 9 May 2016 10:00:59 +1000
- To: Jesse Wilson <jesse@swank.ca>
- Cc: Amos Jeffries <squid3@treenet.co.nz>, HTTP Working Group <ietf-http-wg@w3.org>
On 9 May 2016 at 00:55, Jesse Wilson <jesse@swank.ca> wrote: > Recover immediately on the same connection. (Seems like the definition of > insanity?) This is not necessarily the case. Often, as is with the nginx case, REFUSED_STREAM is highly transitory. All that it takes is a round trip, after which you have their settings updated and you know not to violate their [maximum concurrent streams, flow control window, etc...]. I would try on the same connection at least once. Making new connections runs counter to the spirit of h2. We want a single connection that is good for everything. I would retry exactly once and then consider the request a lost cause.
Received on Monday, 9 May 2016 00:29:15 UTC