"Dynamic range compression" takes all signal above a certain threshold and reduces it by a specified ratio (this is typically called limiting) or takes all signal below a certain threshold and raises it by a specified ratio (this is typically called compression). So let's say you have piece of music where the highest peak is at 0 dB, and the lowest part is at -30 dB. That piece of music would have a dynamic range of 30 dB. Let's say you apply limiting so that the highest peak in the music is now at -10 dB and the lowest peak is at -20 dB. THAT is dynamic range compression. You have changed the music so that there is less of a difference between the highest and the lowest values. There is now 10 dB of dynamic range in your music.
Now, suppose you take that same piece of music and raise every sample by 10 dB, so that the highest peak is once again at 0 dB. You haven't changed the dynamic range of the music. You've simply scaled the volume at every point in the file by a constant. That is normalisation.
Now, let's say you put that file on your iPod and play it. You turn down the pre-cut. Rockbox scales the gain of the file by a constant. There is still 10 dB of dynamic range in the file itself, but every sample in the file is played back at a lower level. The effective signal to noise ratio may be lower, but the dynamic range of the music itself is still 10 dB. The music is not "compressed" by lowering the volume.