i actually tried to do this today just to see if it could be done... the problem with keying the focal length is that it ends up flattening the image to extremes (and zooms like crazy - so i had to pull the camera out and try to match object size as the camera dollys out and the lens zooms in), when the actual orthographic viewports do not... i used a camera with aim and started with a keyframe towards the end of the camera movement, working backwards along the timeline to get the move. i think without keying the focal length it would be fine... just as long as you have the camera along an axis you should get your end result (and then taking mirek's and pm's advice, then going into a 2d compositing program and performing a transition from the camera movement to your still image).