But these Real-Time programmers are more concerned with determinism/predictability as a priority, rather than just performance- that's what I read in the 'Dreams' section.
So they wouldn't like the GC-pauses or hotpsot dynamic compilation pauses, but so long as they're really short, isn't it better for us to have better over-all performance with a small startup time? As far as I know that's the way the client JVM has gone - dynamic hotspot compilation at bottlenecks only & dynamic GC'ing (I imagine that static GC'ing is possible too - ie without a GC thread).
It's interesting how they say that a hardware JVM is possible. It's a one-liner in the 'Dreams' section. That would be awesome, imagine a hardware JVM console - the JBox or the JayStation!