Actually there are two methods to normalize.
There's the peak method and the average loudness (called on some occasions RMS) method.

Most simple programs use the peak method. With this method, the program will scan the audio for the loudest peak, and adjust the volume of the file until that loudest peak is as loud as you wanted.

Let's say you want to normalize to -0.1 dB, then the loudest peak of the signal will top at -0.1 dB

The second method analyzes the average volume of the audio. That's a different concept as the former method. If for example your audio file is pretty quiet, but occasionally there are some very loud peaks, the average loudness will be low (as those peaks are only occasional). So if you set a value for RMS normalizing, this doesn't mean the loudest parts will have that value at maximum. For example if you set the value to -18 dB, that doesn't mean there won't be peaks higher than -18 dB. This should be taken in consideration, clipping can still occur.

What values are considered default?
With peak normalizing, usually you'll use -0.1 or -0.2 dB (FS, so digital dB's). Why not 0? This is a security for when you want to press commercial cd's. In the cd plant, the machine that presses the cd's will see several samples at 0dB as errors. If you peak normalize just below 0dB, the loudest peaks in the signal will never be considered as errors.

With RMS normalizing, it differs from country to country. There are several references. But generally you'll have values between -18 and -12 dB. -10 dB is already pretty loud.

It all depends on the recording of course. With a decent recording, peak normalizing will usually do the trick. With poor recordings, it might get harder. Usually if the overall loudness has to be increased, engineers will generally prefer to compress (up to brickwall limiting as an extreme) the signal, and normalize after it.

good to know, thanks DJ thy.

I usually just normalize to 100% with soundforge. the CDs seem to play fine for me.

I know what compression is, but what is brickwall limiting?

It's an extreme form of limiting. Compression ratio is set to a maximum, and attack is as short as possible. You set a value that the signal can't pass. The limiter will act as a wall. Nothing can pass that certain value you set. As the limiter acts very fast (fast attack time), even transients will be affected, modifying the sound in a distinctive way.

Maximizers often use that kind of limiting to get their characteristic sound.

hmm.. good to know.

i like my stuff to sound as good as possible.. do you think it would be possible to upload some samples of good and bad mastering? I have an FTP you could upload to if you like.. PM me for details.