Many engineers leave about .1 dB of headroom, simply to deal with
older (I mean REALLY older) CD players that would wig out on seeing
a sample value of 32767 (absolute 0 dB).

Most commercial music is way overcompressed and the medium is
pushed to the max.

Mastering is a bit of a "black art" and involves a lot of things,
including setting the peak and volume levels, tonal balance,
saturation or "warmth" and other overall aspects.  Also, an
important part of mastering is to make the songs in a CD fit
nicely together.

There are as many ways to approach it as there are people doing it --
it's definitely more of a "try and tweak" art than a "plug and chug"
science.  Nonetheless, numerical tools are definitely used.

I recommend starting to learn mastering compression using
PeakCompressor.  I can say this with authority since I'm just
beginning to learn mastering and it helps me!  :D

First off, let me suggest to NOT put compression on the master
channel.  Mix your song completely without any master channel
compression, and then mix down to a 32-bit file, and apply any
mastering plugins to that file.  (If I'm doing a single, I
import the mixdown into my song and apply mastering there.
If I'm doing a CD, I use a single new .sng file for all the
tracks, which is very handy for a number of reasons.)

Here are two approaches to compression for mastering.  You'll
need PeakCompressor, and also the
[url=http://www.elementalaudio.com/products/inspector/]Inspector[/url]
plugin.

I generally mix nice and hot to begin with.  This isn't necessarily
the best method, it's just what I like to do.  During mixing, I
avoid clipping on the master channel by adjusting master fader or
individual faders.  If I find I'm having to crank down a lot due
to some parts colliding and causing rogue peaks, or due to some
part that's being nasty, I fix it early on, so I get nice warm
levels all through the recording & mixing process.  (This REALLY
isn't necessary, I just like to do it.)  So when I'm done, I usually
have a mix that isn't quiet as a mouse or anything.  I do mostly
acoustic music, though, so it's not a solid wall of sound or anything.

So when I start to master, I plug Inspector in on the Master channel
(since I'll be using it only to measure, and I'll be using it for all
songs on my CD, and it does no harm).  I put PeakCompressor on the mix
track.

My mixes usually get typical RMS levels of -15dB, a little lower
in the quiter parts and a little higher in the loudest parts.
My peaks are of course already right near zero dB.  For my music,
I find that a 4 dB boost is usually enough so that it sits well
with other music in its genre, albeit maybe a little quieter but
not so you gotta adjust the stereo every time you put the CD in
or take it out (my goal).  A 4 dB boost is very easy to accomplish
with PeakCompressor, so I just set the post-gain on it to 4dB.
I push the Limiter to the top. I leave the ratio at the default
3:1 setting, and adjust the threshold until I only occasional
peaks clipping.  I do this while monitoring to make sure that
the sound remains clear after the adjustment.  Then I draw down
the limiter until my peaks are at zero dB, and not above. 

Finally, I check in Inspector to see what my RMS levels are,
and they're generally near -12dB, going above or below for the
dynamics, but not quite as much as before.

That's not the standard way to adjust a compressor.  Here's the
standard way, and it's really better in general.  But it takes
more passes through to get it right.  (The limiter setting is
pretty much the same, so I won't mention it this time.)

Don't adjust the post-gain first; leave it at zero.  Choose a
ratio you think will help you reach your overall goal.  (If
your music style is relatively natural, the default is a good
starting point.  If you're doing heavy metal or typically
highly compressed material, you may want to start at a higher
ratio, like 4 or 6 to 1.)

Push the threshold all the way up.  Slowly draw it down,
paying careful attention. The sound will get quieter -- adjust
your monitor levels if necessary, or give your ears time to
adjust.  Keep going until you hear unwanted overcompression.
Back off until you're sure the music still sounds good. 
Bingo, you just set your level!  Now push the post-gain up
slowly (and lower your monitor level if you pushed it earlier)
until the peaks are clipping a bit, and adjust the limiter as
before.

See how high you were able to push up the post gain, and compare
that with the before and after RMS levels.  The RMS levels should
go up almost but not quite as much as the gain.  The higher a
ratio you used and the lower the threshold, the less the RMS
will go up for a given gain boost (the bigger the difference
will be).  This isn't really important, but learn to understand
how much gain you need/want, and how that affects the overall
loudness (RMS levels).

Compare your RMS levels with clips you like (e.g., commercial
music, or other indies whose work you admire in your genre).

Most likely, you'll be a bit lower than theirs.  Don't worry
too much -- there's a trend to overcompress in the industry,
for a lot of reasons (most of them "unfortunate").  If you're
within 3 dB, feel great.  If you're 10dB below, that means you're
only half as loud -- probably need to push it more, or your
listeners will be annoyed by having to adjust their volumes when
playing your tracks.

HTH!  I'll have to save this for my upcoming book :)
(Just kidding)