/
What I do not quite understand is the overall assertion that low-level
aspects of OS design can not be simulated. Interrupts were much
discussed, but I ask: can one not simulate interrupts? It seems absurd
to me that this would be impossible.
This is indeed trivial. It can be reduced to the statement that the Python language
is Turing-complete. What remains unclear in this discussion is the perspective
on "Python". Somewhere we have to leave the sim in order to run a concrete
system. The different high level representations of the concept "hardware interrupt"
have to be projected onto one and only one that is feasible by the machine.
If I interpret Your concerns correctly, You obtain a greater flexibility on the
sim-level, which should influence again the "real" OS machine code? The model of this
relationship is Psyco in the PyPy realm: being itself a Python program, that generates
machine code on the fly that drives again the interpreter, that runs Psyco.
But this tangled hierarchy in which OS and Python-Interpreter drive each other may
be my own fantasy, that has nothing to do with Your "prototyping" intention
in the closer sense ... ?
Regards
Kay