This is tricky. I sort of agree with the author and sort of don't. It boils down to: if you own an algorithm f: H->R^n and show m number of (k, v) pairings suc

This is tricky. I sort of agree with the author and sort of don't. It boils down to: if you own an algorithm f: H->R^n and show m number of (k, v) pairings such that f(k) = v, does it follow that you have revealed all or part of the algorithm? (Where user data is folded into f.) This would necessarily have to do with how big m is and whether it is enough to infer a f with reasonable accuracy. Not sure what are good metrics for "reasonable accuracy" though.