Airbag design is not at all like orbit design. You can't just take Newton's Laws and apply them directly, although they are perfectly in action in terms of what causes airbags to work or not. It's just... we don't know to apply them. So what winds up is we just test 'em. We just drop test them, and we see how they fail. And a lot of it is "forensic engineering." We sit there and you look, at what tore. You look at how it tore. You watch the videotapes. You press rewind a gazillion times just to see how it happened, how it failed, where it failed. We put chalk dust on all the rocks so we can correlate exactly, "OK, this rock caused this failure, this rock caused this failure..." and we just sit there and inspect every single seam of the bag, after every drop. And we use our best judgment.

And we have sort of built up this database in our heads and on paper of what different tears look like. We have different classifications of tears. There's the "classical abrasion tear." There's the "bladder tear." There's the "pull-out tear." And there's all these different types of tear and failure modes that we've got catalogued. And whenever we see something that is beyond what we expect, then we sit down and figure out, "Well, why do we think this happened, and what do we think we can try to fix it?" And since we never really know we usually have to try 2 or 3 different things and see which one works better and start going that way. And then try 2 more, 2 or 3 more different things, and keep trying.