I advocate over-regulation. It's not very hard: figure out the exact set of regulation you want, x, and then advocate x&y, where y is not the null set. Since the likelihood of x&y is logically less that of x alone you stand a better chance of getting x implemented.