Philip Martin wrote on Wed, Mar 21, 2012 at 13:00:01 +0000: > Philip Martin <philip.mar...@wandisco.com> writes: > > > So if I have > > > > 0, 1 ... X, Y ... N-1 > > > > where X and Y collide I would expect changing the seed to give: > > > > K, K+1 ... X, Y ... N-1, 0 ... K-1 > > > > Is that going to remove the collision? > > Is it something to do with the bins?. If X and Y were originally in the > same bin the seed change might cause one to move to the next bin so X > and Y no longer collide but Y and Y+1 collide?
The issue discussed on IRC has to do with the hash buckets, yes --- with someone intentionally causing most of the hash elements to be in a small number of buckets, rather than the more even distribution of the 'average case' analysis. > > -- > uberSVN: Apache Subversion Made Easy > http://www.uberSVN.com
