> I wonder if one is willing to change ECN semantics already, by making > CELighter weight than a packetdrop, why not also using an explicit signal for > emergency brake?
This is the principle I proposed with SCE. There, CE remains a broadly drop-equivalent signal (the "emergency brake"), while ECT(1) becomes SCE, a softer and higher-precision signal which is produced in the way DCTCP expects. As of a couple of hours ago, I have three machines in my bedroom which are running SCE-aware Linux kernels, including a "rehabilitated" version of DCTCP which responds appropriately to drops, CE and SCE and is therefore compatible with use on the general Internet. Now I just need to blow the cobwebs off the test harnesses which were used to refine Cake, to ensure that the assertion I just made above is actually true. - Jonathan Morton _______________________________________________ Flent-users mailing list Flent-users@flent.org http://flent.org/mailman/listinfo/flent-users_flent.org