On Wed, Feb 18, 2009 at 11:04 AM, zooko <[email protected]> wrote:
By the way, darcs might also want to consider using the Tiger-192 hash function instead of SHA-256. Tiger-192 takes about 1/3 the CPU cycles of SHA-256 (with most C implementations of those functions, on 64-bit architectures).Whether that difference in speed applies to Haskell, and whether it has any significant effect on darcs performance, remains to be seen. Fortunately darcs has a benchmarking system now which can answer such questions. Regards, Zooko P.S. There is a drawback -- Tiger-192 is more likely than SHA-256 to turn out to be susceptible to collision. However, Tiger-192 is still safer than SHA-1 (used by git, for example), which has already turned out to be susceptible to collision.
I'm not terribly familiar with the hashing, but wouldn't a switch to Tiger entail backwards incompatibility? So we'd've to pay a price for that performance. -- gwern
signature.asc
Description: OpenPGP digital signature
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
