eBPF for cipher .. I dont think it is good idea! Anything apart from offset and above L4 parsing can be costly.
", "time": "2023-11-07T12:19:46Z"}, {"author": "Carsten Rossenhoevel", "text": "Hi @Kireeti Kompella and all, anything using TCP can be \"gamed\". The results of the responsiveness test should not be used for formal benchmarking. We have just released RFC 9411 which describes an exact method of HTTP/HTTPS benchmarking, which puts an immense focus on preventing gaming. It formally addresses network security gateways but the algorithm can be used for any end-to-end performance L7 benchmarking.
\nNOTE that the IESG was puzzled in its review about the clear definition of TCP parameters, which some argued should be left open for individual choice. But such individual choice is the source of manipulations, which convinced the IESG in the end. From my point of view, the responsiveness draft has the focus to evaluate such individual TCP parameters-influenced perceived network quality, whereas RFC 9411 defines an objective, reproducible method for benchmarking.