Re^4: [OT] The statistics of hashing.

[BrowserUk]

I don't have the actual code, but it was something like this.

Thanks for posting that.

However, for my purposes -- deterministically generating a randomly distributed value -- it doesn't work well. With the MD5, every single bit change in the input generates an average of 64-bits change in the output -- which is near perfect.

But this crc algorithm -- which was probably great for your purposes -- only generates a single bit change in the output for each single bit change in the input:

<Reveal this spoiler or all in this thread>

[flexvault]

BrowserUK,

I had to download your code to figure out what you were doing, but now I know your interest in uniqueness was different than mine(i.e. project).

Each of the test encrypted files was unique and we wanted an algorithm that would produce the lowest number of duplicate hashes for the different unique files. The statistics I showed was the number of duplicate hashes produced from unique files. Each hash was generated and then:

  ...
  $Dupes{$hash}++;
  ...

Then we just counted the total number of keys with values greater than one.

Your code was very interesting and quite a mind tester.

Thank you