Rich subject !
Some platforms like Itunes & Youtube apply an average loudness normalizing algorithm, and this is discussed at length in Bob Katz's book iTune Music, even if as the title suggests the book focuses on Apple's system.
The book goes also into details about the very measurable distorsion induced by lossy encoders, when you feed them with high level peak normalized music
I can confirm by ear that all lossy encoders tend to distort a lot when close to 0dBFS. My experience is that a mere -0.5dB on high quality stuff or even -0.3 on more electronic stuff will make encoded files sound much better everywhere, notably when people import the CD onto their computer (thing that everybody seems to be doing nowadays to listen to a mastering....) On very mid heavy music like piano solo I even go below -1dB because the distorsion becomes horrible, and anyway who cares to have a piano as loud than a metallica album ? :-)
I would love to start experimenting uploading stuff on Youtube without excessive peak limiting.
A whole new way of mastering is about to happen...