[typedef_void]

This looks pretty cool.

However, if it's actually useful, why wasn't it invented layer as part of OpenGL / taught in standard graphics courses?

Given that OpenGL already has the 3D model, and has to redraw everything 30 or 60fps anyway, there's no performance hit to slightly change the camera angle.

What are the limitations / side effects of this technique that cause it to not be used in practice?

[vidarh]

The side effect is that people made to endure this more than a couple of minutes as a novelty will want to murder you. They'll then plead temporary insanity and likely get off with a light scolding and a medal.

[jim_dot]

Doing multiple passes with slightly different camera settings is becoming more common, but you would only use it with something like the NVidia 3D system, or any other 3D display technology. You would never use it to do something like this.

[arketyp]

It's kinda annoying.