>>2
There's no need for a GC. Relations between states are representable in timeless (finite, as in the standard model of) arithmetic. If you don't like arithmetic, you can choose combinators or lambda calculus or the turing machine, they are all provably equivalent as far as computability power is concerned (although slightly more powerful than our finite state machines, if we don't allow for increasing memory - contrary to popular belief, all these models of computation allow for unbounded numbers (or in
>>1's case, what he calls BigNums), but they're always finite).
>>1
It might be possible, at least if one doesn't stick to 'plank units' as being the ultimate unit and doesn't want quantization errors (there doesn't seem to be too much evidence for them). At least, IMO, using non-computable reals in physics can be seen as somewhat treacherous given that almost all of them are uncomputable concretely infinite non-enumerable beasts.