> On 29 Apr, 2020, at 12:25 pm, Luca Muscariello <muscarie...@ieee.org> wrote: > > BTW, I hope I made the point about incentives to cheat, and the risks > for unresponsive traffic for L4S when using ECT(1) as a trusted input.
One scenario that I think hasn't been highlighted yet, is the case of a transport which implements 1/p congestion control through CE, but marks itself as a "classic" transport. We don't even have to imagine such a thing; it already exists as DCTCP, so is trivial for a bad (or merely ignorant) actor to implement. Such a flow would squeeze out other traffic that correctly responds to CE with MD, and would not be "caught" by queue protection logic designed to protect the latency of the LL queue (as that has no effect on traffic in the classic queue). It would only be corralled by an AQM which can act to isolate the effects of one flow on others; in this case AF would suffice, but FQ would also work. This hazard already exists today. However, the L4S proposal "legitimises" the use of 1/p congestion control using CE, and the subtlety that marking such traffic with a specific classifier is required for effective congestion control is likely to be lost on people focused entirely on their own throughput, as much of the Internet still is. Using ECT(1) as an output from the network avoids this new hazard, by making it clear that 1/p CC behaviour is only acceptable on signals that unambiguously originate from an AQM which expects and can handle it. The SCE proposal also inserts AF or FQ protection at these nodes, which serves as a prophylactic against the likes of DCTCP being used inappropriately on the Internet. - Jonathan Morton _______________________________________________ Bloat mailing list Bloat@lists.bufferbloat.net https://lists.bufferbloat.net/listinfo/bloat