> > Hence my point: legacy and the rest (as of cpu=v4) are the only two categories > we should have in _this_ version of the standard. > Rest assured we will add new insn in the coming months. > I suggest we figure out conformance groups for future insns at that time. > That would be the time to argue and actually extract value out of discussion. > Retroactive bike shedding is a bike shedding and nothing else. If some existing implementations aren't supporting some of these instructions don't we need a way to make a profile that says that so that tools can know what they have to generate for things to work? That to my mind is the reason we would define the profiles. Sincerely, Watson > > -- > Bpf mailing list > Bpf@xxxxxxxx > https://www.ietf.org/mailman/listinfo/bpf -- Astra mortemque praestare gradatim
