The effect of Meltdown and Spectre in our communities

1) KPTI seems to be a pretty bad performance hit, especially for old CPUs. It would obviously be possible to expand the kernel stub which is still mapped into all processes and reduce the hit that way. Are people looking at what can safely be done in that area? (I strongly assume they are.)

2) Sort of following on from some of the comments reported - "insane that you guys think you can run secrets with non-secrets on the same piece of hardware", the plan that they "most certainly have" - perhaps (we) software - and hardware - people are trying to be too self-important regarding security. We go to great (which may not equate to successful) efforts to keep those exploits we learn about secret until we have produced some sort of broken fix. Would it not be more sensible to spend more efforts educating those users who want to know about the risks associated with software and hardware so that they can prepare back-up plans when something hits, and put more effort into keeping things working when things need patching? The damage an exploit would cause is often (bad) guesswork, but the damage caused by bad patches, and by effort which could have gone into other things instead, is more measurable.

3) Regarding open hardware, is there any current effort into ISAs which are explicitly designed for Qemu-style recompilers? I know Transmeta failed at something like that in its day, but I could imagine that that could make some sense. Qemu is with all respect not a high performance tool, but it would probably still be a lot better with such an ISA. It seems to me that this would be a more realistic target for open CPUs than something like x86, but also not something that the likes of Intel would have an interest in doing in user-visible software.