With Amazon you can switch it on and off and it changes as you listen.
So if I'm listening to a compressed song and have normalization "off" and the volume set to where the quietest parts are just audible and then turn normalization 'on' then those barely audible quiet bits are no longer audible. In other words, normalization seems to have defeated the benefit of compression.
On the other hand, if playing a non-compressed song with the quiet parts just barely audible and I turn normalization off or on, almost nothing happens.
So from an experience standpoint I'm having a hard time understanding what you mean when you say normalization has no effect on compression. I understand it does not change the level of compression or the level of the quiet bits _relative_ to the loud bits. But it does decrease the loudness across the entire DR which lowers the volume of the quiet bits and the loud bits together in which case the quiet bits can become harder to hear.
@glupson, I'm guessing, but do not know with any certainty, that producers could selectively compress specific portions of the DR and not just across the whole range. I would assume this is how artful DR compression is used with classical pieces that have extremely wide DRs.
I would also think that the effect of DR for the listener would vary with various types of music, recordings and production technique.