A discussion with wjl about units in physics led me to this interesting book, titled "Elements for Physics", available free online. It seems to claim to be able to express physical laws in a way that not only respects changes of "linear" choice of unit (e.g. foot vs. meter) as they should anyway, but also in a way that respects changes of "exponential" choice (e.g. foot vs. inverse foot vs. foot-squared). I am quite intrigued, but there's huge gobs of intense differential geometry that are quite intimidating standing between me and understanding. I wish I could see a definition of connection on a manifold that doesn't necessarily have a metric that didn't still go through frickin' nontensorial Γkij coordinates that I have no clue what they mean, really.