Re:Fast and formally verified C implementation of Base64
by kd0amg on Tuesday September 17, 2013 @08:30AM
(#6398858)
Why would that be obvious?
A straight up "no collisions" theorem would require proving limits on the behavior of other objects in and around the road and also that certain mechanical components never fail.
Re:Fast and formally verified C implementation of Base64
by Someone on Tuesday September 17, 2013 @11:17AM
(#6399763)
Again, just as in the base64 program. It is proven correct only in a world without cosmic rays, oxydation, failing capacitors, etc.
The 'only' difference is one of scale (orders of orders of magnitude.)
Problem is, the assumptions that such a model makes will be violated, either because the world is different than modeled or due to failure of a component. That is why you would want additional safety features.
That, IMO, is similar to the case where you build a component for base64 encoding, prove that it never will be called with a null pointer argument, and yet include that null pointer check and a way to signal errors in your component.
You can be sure "It will never happen" _will_ at some moment turn into "that's interesting", but you don't know where or when.