IPv4 Hash Function

Thu Nov 11 09:53:54 UTC 1999

In my experience, Fibonacci hashing is faster and gives a better
spreading.

--vadim

Richard Steenbergen <ras at above.net> wrote:

I have seen the IPv4 hash function:

(ip ^ (ip >> 23) ^ (ip >> 17)) & ~(bucketsize - 1)

for power-of-2 sized hash tables attributed to Van Jacobson, in many
places, but I have never found any paper or writing that is the actual
source of this. Does anyone know where I might find such? Personal
curiosity only.