[rb-general] transitive collision resistance [was: rb formalism]

[Orians, Jeremiah (DTMB)]

> I'm not sure what you mean by 'not possible to collide' here. Hashes are typically smaller than the allowed inputs, which means there must exist different input files that produce the 
> same output hash. A cryptographic hash just makes those collisions hard to find/create, it cannot prevent them.

I am not talking about the same hash, I am talking about different hashes that can't have the same collisions.

> As above, this is impossible to prove because it cannot be true.
For a Single hash but not for all possible sets of hashes given perfect knowledge of the contents of the file

> The hashes will still be smaller than the allowed inputs, which means there must exist inputs that have the same hash.
For particular hash functions, yes that is true

> They may just be hard to find.
Lets assume trivial hashes for a second and finding collisions becomes cheap and easy.
Then we walk through a list of hashes until we find a set which do not share any mutual collisions.

Think infinite number of possible hash functions selected from