The "loudness war" is mainly hype anyway. 0dB(FS) is an absolute limit and as "loud" as you can get--all the compression in the world doesn't let you exceed that. Obviously, what's being talked about the the dynamic range, the difference between the peaks and the quiet bits--and there's no right or wrong to do with this.
What you choose to use depends on all sorts of factors but, in particular, the style of the material being mixed and also the listening conditions. What works for a symphony concert that builds from a solo flute to a whole orchestra crescendo listened to on an expensive audiophile stereo in a quiet living room won't work for a Metallica track listened to on earbuds on a noisy train.
The listening conditions are important here. In the "olden days" when I was younger, music was far from portable and my little record player in my room was always in nice quiet conditions. Nowadays, most music is listened to in cars or on personal stereos in noisy locations. A normal city street can be at around 80dB(SPL) before you even start to add music. There's a pretty small range between "drowned out by traffic" and "hearing damage".
I "mixed" a classical concert a while back (a misnomer for turning on an announce mic a few times) and, bored to tears, spent most of my time watching the SPL meter that constantly reads out at the mix position. Without any music, the audience and things like the ventilation system had the background level in the 1800 seat theatre at about 58dB(SPL). With the very quietest passages of music playing, this was up around 68dB(SPL) and a few crescendos hit just under 100dB(SPL) at the mix position. Many classical CDs reflect close to this sort of dynamic range.
However, most of the live stuff I do is theatre and, throwing in the usual moving lights and rustling candy wrappers in the audience, anything I mix for theatre has to have a significantly smaller dynamic range if the quiet bits are to be heard by everyone. As and aside, a show I did some years back started with a fade in of sound (and fade out of the house lights) that lasted almost two minutes. It was interesting to watch the point at which the audience actually noticed something was going on. For more than the first minute they tended to just talk louder to be heard then, suddenly realised there really was something happening and went quiet.
Anyhow, I digress. As long as music is going to be listened to in cars and on personal stereos, the "loudness war" is already lost--subtlely will be lost even if you try for it. The trick is to fool the ear into THINKING you're getting a variety of levels even if they're within a narrow range of levels.