On Sun, Apr 13, 2025 at 07:59:21PM +0200, Maxime Devos wrote:
> 
> On 13/04/2025 19:47, to...@tuxteam.de wrote:
> > On Sun, Apr 13, 2025 at 07:03:56PM +0200, Maxime Devos wrote:
> > 
> > [...]
> > 
> > > > Typical hashing of the non-cryptographic kind aren't designed to
> > > > virtually eliminate hash collisions [...]
> > Nit: given a "reasonable" hash, the collision probability should be
> > the same for crypto or non-crypto hash (for the same width, for "random"
> > input). [...]
> 
> Many hashes aren't reasonable, then.When looking at documents describing the
> quality of (non-crypto) string hash functions, you can sometimes see
> discussion about how certain hash values are unreachable and how to choose
> parameters such that hashes are more 'spread out' (and hence, making
> collisions less common), trade-offs between speed of hashing and risk of
> collision ...

Oh, there are quite a few "good" non cryptographic hash functions [1].

Of course, it's possible to contrive an input for which they do fail.
Cheers

[1] https://en.wikipedia.org/wiki/Non-cryptographic_hash_function
-- 
t

Attachment: signature.asc
Description: PGP signature

Reply via email to