[strangeattractr]

This is making me reconsider how I do IDs. I thought the performance of sequential IDs was significantly better. So my approach was to use a standard auto-increment primary ID and then obfuscate by id * p mod m where p and m are coprime and very large. then i get back the original ID using the mod inverse. Should I just be using UUID?

[eloff]

I would use uuid in this case. If p and m are too large you get overflow. If they are too small your keys are guessable. If it matters, use uuid and don't waste time and mental energy on it.