flurrything

The secondary hash must primarily be good at mapping the hashed value into a slot: that is, be as cheap as possible (ideally 0 cycles, fib hashing is ~5 cycles) if the primary hash was good while preventing catastrophic performance if the primary hash was bad. All of this without knowing anything about the primary hash.

The whole point of fibonacci hashing is making hash tables more resistant to bad hash functions.

Expecting everybody to be using the perfect hash function for each case is extremely naive, and even then, it just takes somebody else refactoring some code and adding a new member to a struct somewhere without updating the hash function to break that completely.

That link is broken, but if it's the paper I think it is, the key result is that you can extend safe Rust with safe components that do use `unsafe` as long as you are able to prove these components to be sound.

The same is not true of neither C nor assembly, and there are already soundness proofs of most unsafe components in the standard library available, which means that safe Rust + the standard library is sound. With safe Rust + the standard library there are almost no data-structures that you can't write.