i dont actually see why the number of buckets
should be prime - doesnt this depend on the system used
to generate the bucket selection?
Yes. Which should also use primes to reduce the risk of hash collisions. Perhaps you could also distribute your test code, so others could have a go at testing it? Have you tried adding the canonical location of all files on your root drive, for example? I know that dataset in particular took down early Java hashing algorithms (because they where radix based, and thus would generate identical hashes, for nearly identical keys, where as a solid hashing algorithm generates wildly different hashes from similar keys).should be prime - doesnt this depend on the system used
to generate the bucket selection?
its a simple system, this doesnt mean its bad.
I'm not saying it is.