On 6 May 2010 05:52, Sylvester-Bradley, Gareth <Gareth.Sylvester-Bradley@eu.sony.com> wrote:
One question: I think all the hash functions you have so far are *cryptographic* hashes, right? I'd be interested in seeing non-cryptographic hashes like Bob Jenkin's hashlittle2 in the same framework. It looks like that would be possible?
Yes, they're all cryptographic hash functions for now. The concepts would definitely allow models that are not cryptographic. One problem I foresee there is that I didn't have enough models of a message digest to distill a concept, so they currently return an instance of a specific digest template. Because that digest template is algorithm-independent, there is a slight cost involved in the conversion from the internal state. That cost is lost in the noise for the more expensive cryptographic hashes, but might be a problem for the much faster non-cryptographic hashes. I'd love to generalize it further; Do you have a list of non-cryptographic hashes you'd like to see? Thanks, ~ Scott McMurray