Re: Cryptographic hash uniquness (was [Haskell-cafe] Simple network client)

Graham Klyne wrote:

This is a very late response ... but I did some calculations as part of some work I did a while ago: http://www.ietf.org/rfc/rfc2938.txt (See appendix A "The birthday paradox")

#g

A memorable summary of the birthday paradox being : There is a 50% chance of a collision when number of digits in your population size reaches half the number of digits in the largest possible population. e.g. With a 128 bit hash you get a 50% chance of a collision when you've hashed 2^64 files etc. Richard.