The talk/buzz around L4S seems a bit fan-fiction-ey at best. If ECN is un-deployable on the actual internet then what hope do we have on not only building features that require client kernels to reliably pass packet flag data on to clients (extremely good luck with that on Android), but also L4S appears to hope that people will change their routers to support L4S too! It's the double whammy of "hard/unlikely to see real use"
one honk maybe more
toke@social.kernel.o..
replied 15 Dec 2023 18:26 +0000
in reply to: https://benjojo.co.uk/u/benjojo/h/tkvb1xvgXWsRW9GtLl
@benjojo
Unless you have a contractual relationship with the service operator so you can sell them access to your "edge" data centre, where you can guarantee that the markings survive out to the bottleneck access link. In which case being able to deploy an IETF-blessed priority queue at that bottleneck link is a good way to counter any accusations of preferential treatment...
I really hope that's me bring overly cynical, but I could totally see it playing out that way 😬
dwcarder@noc.social
replied 16 Dec 2023 15:47 +0000
in reply to: https://benjojo.co.uk/u/benjojo/h/tkvb1xvgXWsRW9GtLl
@benjojo I’ve figured the use case is for deployment on residential ISPs that have direct connectivity to CDN’s, so it’s more of a limited domain.