[Codel] [PATCH 1/2] codel: Controlled Delay AQM

[Jim Gettys]

On 05/07/2012 02:44 PM, Dave Taht wrote:
> On Mon, May 7, 2012 at 11:32 AM, Jim Gettys <jg at freedesktop.org> wrote:
>> On 05/07/2012 01:52 PM, Dave Taht wrote:
>>> ecn:
>>>
>>> ecn was not in the paper (and it's working well) and I'd like to live
>>> in a world where ecn was the default. I think. Others disagree.
>>>
>>> so perhaps ecn and noecn? And ecn be the default (after some serious testing?)
>>>
>> I think noecn *must* be the default at the moment.  There is still too
>> much brokenness out there (though it is improving); Steve Bauer can
>> enlighten us how much.
> Possibly influenced by how hard it is to cycle builds on embedded
> hardware and test one thing at a time,
> as well as the space program's major success at shortening the path to
> the moon, I am an advocate of all-up-testing.
>
> http://www.nasa.gov/offices/oce/appel/ask-academy/issues/ask-oce/AO_1-7_F_snapshot.html
>
> Going with ecn enabled by default, and being aware that it might cause
> problems, will
> result in less time being spent testing each individually.
>
> I note that on an ingress qdisc ecn has mildly less risk as it only
> tests the router and host device.
>
> On egress, yes, you're testing the whole internet.

You want to turn it on in CeroWrt: sure.  Expert audience, who may be
able to notice when things go sour, fine....

Neither you, nor I are expert at the current state of ECN in the global
Internet: but Steve Bauer is.  We can have that conversation with him
shortly.

So for now "First, do no harm" should be our mantra....

So the upstream patch default state should be off.  We don't want to mix
"codel doesn't work" with "ecn is busted someplace", as it gets its trials.

                    - Jim