I have read with interest the description of Highlight Tone Priority for the Canon 5D Mark II and it seems to be in direct conflict with an emerging theory about "exposing for the highlights". This idea states that the raw digital file is a linear file such that there are six "zones" of exposure with the following characteristics: five stops down:64 levels of light four stops down:128 levels of light three stops down:256 levels of light two stops down:512 levels of light one stop down:1024 levels of light highlights:2048 levels of light They conclude that we should "shoot toward the light" because the highest light areas have the most "data". I am having a hard time accepting this "new" theory about exposure. Any ideas?
It's not a new theory, it's been around for years.
It has a grain of truth in it, but you expose for the right exposure, not for some digital signal level. You want to make sure the highlight don't clip. Once you've clipped a highlight, you've lost any data that was in it. So you can "expose to the right", meaning that you probably don't want your histogram clustered on the left (unless you're shoting a black cat in a coal cellar) - though you probably don't want it all clustered on the right either (unless you're shooting a polar bear on snow).
HTP can help in the case of images with extreme dynamic range since it will effectively reduce the exposure for highlights, preventing them from being clipped. HTP has limited capabilities of course. It can only correct by about 1 stop. The downside is possibly slighly high noise in the shadows.