Wiki article on normalization, Please help, especially re: clipping Options

[AndyH-ha]

Normalization is a change to the data. As long as new peaks do not exceed 0dBfs, there is no damage, artifact, or whatever dire consequence some might fear (other than the potential for clipping when played through inadequate DACs, as described above). The audio will just be louder -- which is one of the most significant changes possible, psycho acoustically speaking.

Psycho acoustically speaking, the change can be reversed if you know the amplification factor used, but physically there will be small, permanent (unreversible) changes in the data. These changes will generally be near the lower limit of possibility (i.e. somewhere near -96dB down in 16 bit data) but if you want to be very strict about it, the process is not "non-lossy".

[AndyH-ha]

I don't know what "help" you might expect with lossy encoders but previous discussions have been about the potential for clipping in mp3 that did not exist in the source when peaks are at or very near 0dBfs.

[sheh]

Help in retaining more low signals, which may otherwise be deemed inaudible by the encoder and dismissed entirely. But this is all just conjecture. For all I know encoders might anticipate playback at increased volumes and take that into consideration.