At 19:06 14/05/97 +0000, John Payson wrote:
>>  Hello,
>>  I've found that performing a XOR operation with all the un-inisialized
>> values of the user RAM at startup (in the PIC16C84), you get a good random
>> seed.
>
>Yes, but if two PICs happen to XOR to the same value you're sunk.  Even
>using a 2-byte XOR of user RAM is no guarantee of safety; depending upon
>the characteristics of the chips you may find that certain chips tend
>toward the same stuff.  For a CSDMA/CD application, that would be a
>killer.

I'm not familiar with all the details of what you're doing, but wouldn't it
help to include some kind of a serial number in your code (that changes
with every device), and use this, maybe together with other "randomizing"
measures? Wouldn't you need some kind of individual id anyway? So just
include it in your seed calculations, and you (almost) never get two identical.