Mice worked pretty well in 1992. They worked well in 1984 on the Mac, and five years before that, on Xerox hardware and LMs (though they suffered from "really small ball bearing" disease, and easily got dirty or cranky and refused to roll well).

With a mouse, you have at least one button you can signal an event with.

Imagine doing a UI where you didn't have a mouse button. All you can do is move and point. That's a Kinect, for the most part.

I haven't used a LeapMotion, but I suspect it's the same problem; there's no way to generate a discrete event. It's all fuzzy. Did your fingers touch? Did you wave in a particular way? Some fuzzy matcher is pumping out "90% probability of event X, 75% probability of event Y" every few milliseconds, and it's up to higher layers to turn this goo into decisions that people are happy with. It's hard at all layers.

I really think you need a button, a clicker. Something "hard" in the UI that slams a voice of reason into that fuzzy tower that's continually only able to /guess/ what you're trying to do.

[We wanted a clicker on Kinect. Politically impossible. I think it would have helped a lot.]