Luckily I still have a non-smart TV. If I have to get a new one I suppose I can either refuse to allow it into the LAN, or segment it off somehow onto a second, untrusted LAN, so as to stop it from getting near my other boxes or even attacking LAN infrastructure.
There is definitely going to be some serious arse-kicking at Intel, I would hope. What were they thinking? I wonder about sneaky unchecked writes to supervisor space?
I know all about what speculative execution is, even though I am a very old machine code programmer, but I would very much like to see some of the Dutch PhD student’s code to give me a concrete example.
I am hoping that Intel will put in a 'bug fixed' flag in CPUID, so that operating systems can skip all the wasteful nonsense that they are having to put in just now. I wonder if they have already made such a definition now ahead of time, so o/s's won't have to be re-released yet again merely to pick up the CPUID awareness thing. It would also be nice if we could have a boot option to opt out of this mod for the special case where if we only ever have trusted code in our boxen. And I'm assuming the o/s designers will put in a different code path, or better a different build altogether, for AMD CPUs right now.
If it is true that Apple has already fixed the bug (when?) then perhaps the performance impact isn't quite as bad as asserted (seeing as no-one has complained), or perhaps the hit is only as bad as 17%-30% or whatever if you have a peculiar app that does a huge number of ring transitions to kernel mode.