Re: [tsvwg] plan for L4S issue #29

[Sebastian Moeller <[e-mail]>]

Hi Mikael,

these are good points, that could be actionable for team L4S, see below.


> On Sep 30, 2020, at 13:06, Mikael Abrahamsson <swmike=[e-mail]> wrote:
> 
> On Wed, 30 Sep 2020, [e-mail] wrote:
> 
>> players may be useful too (e.g., I don't know whether an access router
>> vendor offers an L4S implementation, and availability of applications or
>> content expected to benefit from L4S transport certainly would be very
>> helpful).
> 
> I participated in FQ_CODEL / CAKE evaluation which was easy to do because it was developed using a fork of OpenWrt which had readily available platforms/code/compilation environment etc and it was easy to get started because it was already packaged. There was also tools like Flent to evaluate performance in different combinations. All development was done on github etc and it was easy to follow.
> 
> When searching for L4S patches I find https://github.com/L4STeam/linux
> 
> I don't see much activity in the past 5 months, is this really where development is being done?

	[SM] If the repeated claim that team L4S invites everybody to help improve L4S it would be a great idea to offer a set of globally diverse distributed L4S-enabled endpoints, that sit behind well described L4S-AQM bottlenecks and that allow say something like long duration speedtests with timestamps, so that everybody can use the same set of testing endpoints to finally get to the answer of the question to what degree the current L4S design is actually going to work over the existing internet. If applying measurement endpoints is too big an ask, at least team L4S should consider to supply ready configured container/VM images for endpoint and AQM roles to expedite testing.
	Which accidentally does not require any progress for the L4S drafts, and could be tackled today. As it stands, everybody wanting to kick the tires of L4S is basically forced to establish both endpoints and the AQM themselves, which does put considerable hurdles into the path of non-full time AQM researchers. 

Best Regards
	Sebastian

> 
> -- 
> Mikael Abrahamsson    email: [e-mail]
>