[aprssig] Aprx v2 beta, now with Viscous Digipeater
scott at opentrac.org
Thu Oct 22 22:00:42 EDT 2009
> I did some studies over various "rapid hash algorithms", how well
> they distributed their results - of what I had been recommended
> to use by other people, worst ones did result in about 25% of
> value space getting any hits at all, and even smaller fraction
> getting most hits. As a young man then, that did tech me not to
> trust hearsay, and want to do validation with _my_ data.
The T2 has a fixed number of slots, and the first available one (or
oldest if the list is full) is taken. I'm using CRC16. Assuming
perfect hashing, the chance of two 16-bit hashes in a set of 15 (the
default T2 dupe list size) colliding is about 0.16%.
I considered that when I coded it, and decided that was an acceptable
rate for falsely dropping packets as dupes.
> There are very few "good" hash algorthms, and they tend to be called
> "Cryptographic Hash XYZ". But your first task on storage constrained
A cryptographic hash has the additional property of being very hard to
deliberately generate a collision for. You don't need that when
someone's not deliberately trying to screw up your hashes.
More information about the aprssig