@dalias @rysiek Oh, sure. I absolutely believe Tannenbaum was right and every year that ticks past is proving that. But I also think "we have to rely on the magic assumed infallibility of MMU hardware to build security boundaries" is just an utterly dismal cop-out when we also have the ability to build those boundaries in software too. We can have good things, actually.
And in the case of Linux, which his already gone down the route of ring0 drivers, the problem absolutely *is* the language just as much as it is the lack of userland drivers.