Re: [tsvwg] Comments on L4Sops [was Adoption call for draft-white-tsvwg-l4sops ...]

[Jonathan Morton <[e-mail]>]

> On 16 Mar, 2021, at 1:06 pm, Pete Heist <[e-mail]> wrote:
> 
> [PH] I think Jon was referring here to end users as opposed to network
> operators. End users are "innocent bystanders" because they may be
> using a 3168 AQM without knowing it.

Not "as opposed to", but "as well as".  It is the end users who will first notice the problem, to be sure, but the burden of mitigating the problem falls on the network operators, since the end user cannot do anything to evade the problem themselves.  In many cases they will not even observe the L4S traffic, which is causing the problem at some intermediate mutual bottleneck on their intersecting, but otherwise independent paths.

> Examples:
> 
> * Their ISP uses fq_codel in its network.
> * They bought a WiFi AP that has fq_codel in the driver.
> * They're using a public WiFi AP that has fq_codel in the driver.
> * They enabled an AQM in their home router.
> 
> The draft can't immediately help those users, because it will take
> years for the existing AQMs to be updated, reconfigured or undeployed.

Correct, though I would also include the four types of network node above, and the people responsible for their installation or firmware updates, as "innocent bystanders".  They have a reasonable expectation that new protocols or algorithms, which they have not opted into in any way, are compatible with the existing RFCs as commonly implemented and deployed, and which hitherto have been associated with a noticeable improvement in service quality.

The fq_codel AQM explicitly conforms to at least three distinct but related RFCs: RFC-3168, RFC-8289, RFC-8290.  None of these contemplate the need for an AQM to treat ECT(1) traffic differently from ECT(0) traffic.  Indeed, the final version of RFC-3168 explicitly *added* handling ECT(1) equivalently to ECT(0), relative to an earlier version of ECN, and there is discussion of the incremental deployment of middleboxes that respect that, phasing out the older spec.  (That was 20 years ago, so can reasonably be expected to have been completed to within epsilon by now.)

 - Jonathan Morton