Re: Is a faithful HTTP/2 response scheduler necessarily O(n) in the worst case? from Cory Benfield on 2017-01-25 (ietf-http-wg@w3.org from January to March 2017)

From: Cory Benfield <cory@lukasa.co.uk>
Date: Wed, 25 Jan 2017 08:32:27 +0000
To: Tatsuhiro Tsujikawa <tatsuhiro.t@gmail.com>
Cc: Tom Bergan <tombergan@chromium.org>, Stefan Eissing <stefan.eissing@greenbytes.de>, "Kazu Yamamoto (山本和彦)" <kazu@iij.ad.jp>, HTTP Working Group <ietf-http-wg@w3.org>
Message-Id: <C2828755-7067-49B0-8FFE-A417B265FDC7@lukasa.co.uk>

> On 25 Jan 2017, at 02:01, Tatsuhiro Tsujikawa <tatsuhiro.t@gmail.com> wrote:
> 
> If large n is concern, server can limit SETTINGS_MAX_CONCURRENT_STREAMS to the reasonable number.  Usually we use 100, and it is pretty fast.=

The various Python HTTP/2 implementations also set SETTINGS_MAX_CONCURRENT_STREAMS to 100.

Cory

Received on Wednesday, 25 January 2017 08:33:03 UTC