W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2012

Re: Benjamin Carlyle http 2.0 expression of interest

From: Patrick McManus <pmcmanus@mozilla.com>
Date: Mon, 23 Jul 2012 19:17:30 -0400
Message-ID: <1343085450.2859.13.camel@ds9>
To: William Chan (陈智昌) <willchan@chromium.org>
Cc: Amos Jeffries <squid3@treenet.co.nz>, ietf-http-wg@w3.org
firefox does use idle ping for K-A, at about 50s intervals (60s seems to
be a hot button number with nats)- but we close the session just before
we would need the 3rd one (assuming no other non-ping use of the
session), as 3 minutes has generally been the point of diminishing
returns for an idle connection to be reused anyhow.

its a tradeoff.. using it as a k-a means we have a higher degree of
confidence in the connection when we go to send a request on it (and we
avoid taking the delay of a failed liveness check done in realtime with
the request), but on the downside we need to close the connections at
some relatively early point so they don't just sit there chewing up
battery.

-P

On Mon, 2012-07-23 at 16:05 -0700, William Chan (陈智昌) wrote:
> Chromium does not use SPDY PING as a keepalive mechanism. That just
> unnecessarily wakes up the radio on mobile devices. We send PING when
> we have not sent anything over the SPDY session in awhile and we're
> about to send a SYN_STREAM frame. This is a liveness check, which is
> very different from a keepalive. Therefore, we do not have metrics
> about "how many PING are actually needed to keep-alive" since we don't
> actually try to do keepalive.
Received on Monday, 23 July 2012 23:17:58 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 23 July 2012 23:17:59 GMT