I've seen several articles about "the loudness war" and over use of compression and how low dynamic range ruins music, but looking at the dynamic range measurements of some of the music in my collection, I have some tracks/albums with nice high dymamic range measurements that sound very nice, however, there are also some that have very low measurements, that also sound good. How does that work?
There seems to be some meandering in this thread, but the answer seems to have appeared, even if it is a bit technical rather than poetic.
DR = a crest factor. Difference between a the bulk perceived
loudness and the transients.
It tells you nothing about the tonality or clipping distortion (although could hint, along with the rest of the stats the DR meter generates). It's a blunt tool which had it's heart in the right place.
The 'live' version, a plugin that sits in your audio editing program (DAW) is arguably more useful for the engineers among us creating or engineering the material. Gives an idea of just how much transients are being squished (if there are any).
Things have mostly moved on now.
We tend to use Loudness standards (EBU-R128 / ITU-R BS.1770) which allow us to determine the light and shade and actual audibly perceived dynamics of the program material. Some weighted curves and various gating and window measurements form a set of results.
We could scan a track using loudness tools and get the integrated loudness (I-LUFS) which is a bit like a better version average VU of the entire track.
We'd get upper and lower threshold limits and a difference between these showing how loud and quiet those respective passages are - if they are very close and close to the Integrated value, you know it's going to be a fairly homogenous sound in terms of constant volume/loudness.
True Peak is also something that has been tagged with the loudness standards. In simple terms it just up-samples the audio, sees what new peak values are created from doing that and therefor gives an idea what would happen when the digital signal is converted to analogue. Digital audio is not a stair step, it's just samples in a continuous wave. That wave can go above 0 dBFS. True peak would show that, with values above 0 dBFS.
Note: I mention perceived loudness in relation to DR meter. It wasn't really accurate. It was just a weighted and gated RMS value.
Also note: VU meters of old showed music loudness in terms of Volume Units. This was a rudimentary way of showing momentary loudness of audio. Could be skewed easily with lots of bass energy, which we don't hear as all that loud, but the meters would be pegged. Likewise something loud and very bright, but without much bass energy wouldn't look so loud on a VU meter, but my goodness we'd know we had to turn the amp down!
"Classical" music typically has higher transient information, as well as a bigger loudness range (loud to quiet passages) than produced music. Studios use all manor of techniques to shape a sound, "classical" avoids all that, trying to be as pure a recording as possible. Generally speaking.
Grab a 'greatest hits' type of "classical" CD and see how much louder the tracks (movements, pieces) are compared to the source it was taken from, such as a CD of the full opera. In Mastering the CD for the general
unwashed masses (of which I'll class myself one of) they bring the level up using dynamic compression and expansion and limiting, so that the listener doesn't have to crank their amp to hear the quiet bits. They'll also make a more consistent loudness between tracks, and maybe EQ some to make them seem like a more homogenous collection.
Even lossy audio, like MP3, can be of no noticeable impact if at 320kbps constant bitrate or - even better - at V0 variable bitrate. The latter statement is true regarding most music productions but certainly does exclude some very complex recordings, like those that can be found in the 'classical music' genres. Everyone can simply very this, e.g., using foobar2000 with the 
