Of course, that's a valid point. But again, lots of (perhaps most?) people engage in conversation for several hours per day, and a lot of that time is spent gesturing, mostly unconsciously.

I'm not envisioning it as a primary input device, naturally (although the deaf seem to get by fine). But I don't think I've ever been in a situation where a heated conversation made my arms tired; as a replacement for a few basic mouse tasks, I bet it could even improve ergonomics.

Just try it: raising your arms as if gesturing conversationally is way, way less stressful than reaching out to touch your monitor.

Right now I'm sitting back in my chair, with my elbows on my chair's armrests. With comfortable armests I reckon I could gesture all day. If I had a hundred odd inches of monitor surrounding me like in Minority Report it might be a reasonable way to control things -- easier than moving a mouse pointer all the way from one side of a vast piece of screen real estate to the other.

I'd love such an interface precisely for the ergonomics. The mouse is very bad for my wrists. I imagine waving my hand through the air wouldn't be.

Also, the gorilla arm would go away fairly quickly. This used to make me tired:

http://www.youtube.com/watch?v=mPKbCsWLe4k

Now it doesn't. I expect people would get used to minority report interfaces, particularly if they didn't immediately try to transition from 8 hours of keyboard to 8 hours of minority report.

My real question: is source code for this demo available?

I don't think it's absurd to use this for hours a day, given that for many deaf people signing is their only way to communicate face to face, and people will find ways to talk all day.

I work in construction, I spend my days holding 5lb drills in front of me. Using the standard gesture position (forearm down) your arm won't get tired during a day. I basically live my day in the 'gesture position' often for well over 8 hours a day, always with weights in them.

When the upper arm is lifted is when your arms get not so much tired but literally drained, which is what I think the Kinect will likely have to adapt to.

So you're taking it as assumed that there is one, singular interface that people will be using for the whole day. Even now, while designing and thinking things through I usually alternate between typing in a text editor, drawing on a whiteboard, taking notes on paper, discussing verbally with other people in the office, and just pacing up and down thinking to myself. The whiteboard for example is essentially a specimen of the much-maligned vertical touchscreen interface, and I don't see why it couldn't be made into a computationally-driven dynamic display.