Hi Sandy, On Sat, Apr 2, 2022 at 6:45 AM Sandy Harris <sandyinchina@xxxxxxxxx> wrote: > > In switching to using a hash function, this required us to > > specifically initialize it to some specific state, > > Hash functions do not require that. Any such function must > work correctly with a new input block and a more-or-less > random state from hashing previous blocks. Well yes and no. Strictly no in the sense that blake2s_state has a few book-keeping variables, which we probably benefit in terms of caching from having next to the other state variables. Almost yes in the sense that in the ideal model, the hash state can become _anything_ so initializing it to random might be okay. But in practice, maybe not, because at the moment the latent entropy plugin is actually expanding a 64-bit seed with a basic LFSR, rather than supplying more uniformly random bytes (I have a patch out for that now). These details might matter, so rather than tempting fate, just calling blake2s_update the way the hash function is intended to be used seems a lot more cautious than poking at the function's innards unnecessarily. Jason
