Ok, we all know about compression/limiting, which is often used in mastering to push up the peaks. It's done with varying levels of skill, to make the music sound louder.
Normalization is limiting to unity gain -- lifting the peaks until they hit 0 db, but not using any compression.
But ... why don't consumers have control over the range of volume. I know that I have to sift a lot of good content out of my classical playlist during a dinner party, because the dynamics are too wide. If you can hear the majority of the content, the peaks will deafen you.
So, why not de-compress the music, and leave the dynamic range up to the listener. How do we do this, you ask? Well, the technique is simple. After recording a track or album, you first do standard normalization. Then you cancel out the noise, and try to make the silence as silent as possible. So far, these are standard mastering steps. The next part is the interesting stuff.
After you've finished all your standard mastering, you pick a level that is as quiet as your music ever gets. Be careful not to chop anything, but you want to get as high above the base level as you can. Then, we expand everything. Anything that falls below your selected level is made silent, and the remaining range is expanded to fill the 96 db space.
If you've recorded and mixed at 24 bit and you do this step along with the conversion to 16 bit, you shouldn't lose any data precision (and in fact may gain a lot, because rather than having to squeeze 2e24 values into 2e16 values, you may only be squeezing 2e22 values in there).
Of course, if you leave your music like this, a lot of it will be barely audible, with amazingly loud peaks. What it really requires is another knob on your stereo. Call it "Range" or something. What this range knob does is take the digital signal, and apply something like the following formula to each sample:
level / ((1/range) * 96 + 1) - 1
What this does is allow you to digitally compress everything while you listen. You can press everything up as loud as you want, or you can leave it open, for wide dynamic ranges.
Your volume knob will set the peak level, and the range knob will set the relative volume of the minimum level.
Because, face it, sometimes you're listening to your music and you want the range, but other times you want it in the background, and the range is only distracting.
Of course, you don't want to do this to individual tracks necessarily. Depending on how cohesive your album is, you may want to use a larger sample. A good example is an opera, where you want to keep the transformation the same across 3 or more CDs.
With the standard CD format, this is pretty much all you can do (I think), but if the original minimum level could be stored in the audio format, turning on manual compression could be optional. Leave it off most of the time and each CD or track will use its default value to compress as necessary, but turn it on to be able to flatten out the music however you want.