I think Linus' argument is if the code is so hard to follow you can't see
how it ended up in it's final state then that's a problem with the code. The
worry is using debuggers can short circuit understanding the code path and
result in fixes that address the symptom rather than the cause.
I can see the logic although I think he discounts the usefulness of being
able to diagnose system state at failure points. Certainly I'd hate to debug
my user-space code with just a faulting address and register dump at the