Extensive control of program flow might be very difficult to program and quite cumbersome. exactly. But to my mind this is the big point. (Although i am very often wrong) It seems to me, that the problem lies in the function of HW key in
program. If it is used in "check the existance" way then you can easily remove the checks from binary code. And it does not matter what is the essence of checking- You will always have CMP KNOWN_DATA, HW_KEY RESPONSE. that makes me sad. If you are planning to use RND generator then here is the weak point- it only takes some time to locate it (even physical one ) and in case of everybody-reads-everything-and-writes-too situation you could feed this program what uses HW signatures with known data. And the program will never know the difference.
Another thing - how practical is this hardware? If it is implemented on a micro-controller then it can be disassembled is the code inferred via other
OH, I have not given it a really good thought. ucontroller seems to work fine - since for obvious reasons you cannot put there 2^32 bits of ROM. I have used MC68HC705 with printer ports. But of cource you must concider the time it takes and breaks.( And maybe it is better to use some Unix system to begin with where root must be the 'responsible one' with license servers.) JP.