> What would the best possible version of this policy look like if it were smart?

It depends extremely on what your actual goals are, but if we suppose the goal is to increase US manufacturing capacity and economic independence with minimal pain, you'd enact this on a time-table over a period of years, after building enough support across the political spectrum to make it look like this policy will be essentially, if perhaps with some occasional tweaks, be maintained across administrations for some time. You'd also probably not significantly tariff the whole world, at any point.

The biggest problem with this approach is investor confidence. It's hard to sign on the dotted line to spend billions building factories when it's difficult to say with any certainty that these tariffs will still be around by the end of the year, let alone in five years. This was doomed to be way more painful than necessary when he started flip-flopping on this over the last couple months, regardless of whether it's even fundamentally any sort of a good idea.

If the goal's to replace large portions of income tax, as has been suggested by some including by Trump, uh... the math just don't work even with a drastically smaller federal government, it's nonsense, there's no "best possible" version there.