Whenever you apply a Curves or Levels tweak (or even a Hue/Saturation adjustment) to an image, you're degrading it a little by throwing away some image data. Once that data is gone, you can't get it back. In the digital darkroom, this degradation is no longer an issue because you make edits to layers above the image rather than to the image itself.
If you're a photographer, think of it this way: Your unedited image is like a negative that you can print through many different filter pack combinations on many different contrast grades of paper. You can make huge changes from print to print, but the negative itself doesn't change. If the unedited image is analogous to a negative, adjustment layers are like enlarger filter packs on steroids. You can change the color balance as you would with a filter pack, but you can also change the contrast and do local, selective editing akin to dodging and burning. However, unlike their analog counterparts, you can always undo digital dodging and burning.
There are several other reasons why we love working in Photoshop's digital darkroom.
You can use adjustment layers as effectively on CMYK or Lab images as on RGB (though we still typically work with RGB images when we can). If you prefer to work by the numbers, the Info palette shows before-and-after values while you're working an adjustment layer's controls, and you can place color samplers to track key values just as you can on a flat file. If you'd rather work visually, you can use Proof Colors to see how your edits will work on the printed result. In fact, Photoshop even lets you preview the individual CMYK plates while working in RGB.
Adjustment layers and high-bit images
One of the biggest changes in Photoshop CS was the ability to use layers, including adjustment layers, on high-bit files. We love this capability and use it extensively, but we'd be remiss if we didn't point out that doing so can create extremely large filesthink gigabytes rather than megabytes.
In Chapter 11, Building a Digital Workflow, we discuss strategies for dealing with ballooning file sizes in much more detail. For now, a relatively simple rule of thumb is to make your biggest corrections on the high-bit data. After that, we suggest staying in high-bit mode until it hurts, or until you need to do something that you still can't do in high-bit mode, such as running Extract, or some of the more esoteric filters.
The techniques in this chapter work in both high-bit and 8-bit modes. You need to make your own call on when to downsample to 8 bits per channel based on the quality needs of the job at hand, how much time you're willing to spend on an image, and the capabilities of your hardware. You may elect to downsample to 8 bits right in the camera by shooting JPEG, or you may decide to preserve your high-bit data all the way to the output process. However, unless you're shooting JPEG (in which case, you have no high-bit data to start from), we strongly recommend that you make your big initial edits in high-bit mode, and save the high-bit file. That way, if you run into the wall in 8-bit mode, you'll still have a fallback position.