[Make-wifi-fast] [Cake] gaming dscp codepoint?
dpreed at deepplum.com
dpreed at deepplum.com
Wed Jul 25 14:54:03 EDT 2018
I don't quite understand this. Diffserv may exist, but no Internet operators support it. It may be "supported" in edge devices, but since the "service" is in the network queueing, who cares?
The Linux gamer market is rounding error in the global game community.
And console gamers depend on the ISP's non-implementation.
Caveat: I have said for years that diffserv's huge list of codepoints is essentially the result of a committee that has gone wild, creating a standard that is missing any useful path to adoption. There's no computable "translation" of the vague descriptions in the standard to a predictable router queueing behavior. This is super true in places like LTE, where you can observe bad congestion and bufferbloat even today.
Discussing what codepoint means what is like discussing trivia about an imaginary fantasy land.
From: "Dave Taht" <dave.taht at gmail.com>
Sent: Tuesday, July 24, 2018 8:34pm
To: "Jonathan Morton" <chromatix99 at gmail.com>
Cc: "Cake List" <cake at lists.bufferbloat.net>, "Make-Wifi-fast" <make-wifi-fast at lists.bufferbloat.net>
Subject: Re: [Make-wifi-fast] [Cake] gaming dscp codepoint?
On Tue, Jul 24, 2018 at 5:06 PM Jonathan Morton <chromatix99 at gmail.com> wrote:
> > On 25 Jul, 2018, at 3:01 am, Dave Taht <dave.taht at gmail.com> wrote:
> > cs4?
> > cs4 and cs5 end up (typically) in the (oft buggy) linux vi queue on wifi.
> > ef?
> > do any of the gamers here observe any codepoints in use? When I
> > surveyed this a few years ago, I saw very little usage, and what
> > little there was was all over the map.
> If it were up to me,
Wasn't my question. Do you observe any of your games using any codepoints?
>I would use EF for realtime position/command updates and voice comms, CS0 for everything non-time-critical (like matchmaking, garage, shop), and CS1 for downloading patches.
Well, I was leaning towards cs4. I no longer remember what ef maps
into on linux wifi, particularly since qos_map_set was created for
hostapd. Is openwrt tweaking that at all?
These days my aps do not use 802.11e at all and I'm about to push a
change forcing a max of 2ms per AC via the beacon.
Clients on campus don't seem to use much dscp but I'm certainly seeing
ecn from apple devices now.
> Reason is, EF is the only DSCP I can count on being interpreted as "latency sensitive" rather than "for video streaming".
> - Jonathan Morton
CEO, TekLibre, LLC
Make-wifi-fast mailing list
Make-wifi-fast at lists.bufferbloat.net
Reed Online Ltd is a company registered in England and Wales. Company
Registration Number: 6317279.Registered Office: Academy Court, 94 Chancery
Lane, London WC2A 1DT.
More information about the Make-wifi-fast