Re: [tsvwg] Fwd: New Version Notification for draft-heist-tsvwg-ecn-deployment-observations-00.txt

[Jonathan Morton <[e-mail]>]

> On 19 Feb, 2021, at 11:36 pm, Bob Briscoe <[e-mail]> wrote:
> 
>>> Given you have close contact with this ISP, have you asked, or could you ask them whether they have ever deployed any FIFO ECN-enabled AQMs themselves?

>> As far as I'm aware, they have only used fq_codel for ECN.  Before that, they were using SFQ in some places, without any AQM.
> 
> [BB] So do you suspect that any of the CE-marking in traffic not traversing their backhaul FQ-CoDel boxes is from a FIFO ECN-enabled AQM?
> 
> Or is this ISP particularly attractive to users into free and open-source software, in which case I would have thought the CE marking is most likely to be coming from FQ-CoDel? Or can we just not really tell?

They're a co-operative ISP, which means they have a smallish hardware budget and want to make the most of it in terms of utility to the customer.  That is why they've deployed fq_codel in the places where that's feasible so far.

We have no insight at present into the nature of observed ECN activity from elsewhere, except for an observation that some ECN traffic is carried through tunnels.  We do consider it *likely*, however, that most of it is flow-aware AQM of some description, simply because practical AQM deployment has been driven by the ease of fq_codel configuration.

> As the rest of your email changed subject to tunnels, I've split that into a separate thread.

I mentioned tunnels because I infer your aim is to discredit the argument, made more than a year ago, that legacy AQMs may exist that don't also perform flow isolation, and that this is problematic for the L4S safety case.  Encrypted tunnels that intentionally hide the distinction between flows carried within them exhibit similar behaviour, and it's easier to show that they exist in significant numbers.

There is no need to make a new thread about that unless you have something genuinely new to say about tunnels.

 - Jonathan Morton