John Gilmore points out that in hardware, "proving anything becomes very slippery..." Seems to me that we can establish degrees of confidence, in much the same way as pertains to social sciences research: probabilities that a given result is accurate. So for instance, we could say that a device made with parts purchased randomly for cash over the counter at a number of different suppliers, might have a lower probability of compromise than one made with parts mail-ordered from the same large supplier to the producer's workshop every time. Over time it may be possible to quantify the measure of confidence. -gg.