Re: [hash2] Difference from paper: HashAlgorithm interface vs. function object

From: Peter Dimov (pdimov_at_[hidden])
Date: 2024-12-12 02:06:03

Ivan Matek wrote:
> Can you please elaborate why is this necessary(not speaking generally, just in
> this particular case)?
> I originally though we do
>     Hash h2( h );
> so that seed of h matters, but  then I noticed that anyway at the end of
> function we do
>     hash2::hash_append( h, f, w );
>     hash2::hash_append_size( h, f, m );
> so for I presume seed of h matters even if we default constructed hashers in
> the loop.

It matters for the final hash value, but it won't matter for creating collisions
by crafting unordered maps.

That is, any collision will be seed-independent because w will not depend on
the seed. And seed-independent collisions are bad.