I am in the "I can only pick 'em at 192 or lower" category and I find it a little disconcerting that some want the maximum linearity you get with FLAC/WAV anyway.
The linearity present is due to the way we want preserve the signal, without realising that this is an interpretation of the original sound in the first place.
If the original sound is a live acoustic environment, the placement of the mic is what is represented by ultra-linearity. Stereo micing is an attempt to capture a "I am in the room" experience to bring to your ears, so spatial positioning is important, however much can go wrong here in terms of getting the signal to perfectly match someone sitting in the room sitting perfectly still in exactly one spot - only when the recording has those properties covered does the question of ultra-linearity become important. Some may be aware that losing linearity is not necessarily a less true reproduction of the original, but they just like it linear anyway.
I don't like ultra linear, I know if I was in the room, I wouldn't be sitting perfectly still, I would be getting different ambient reflections whenever I moved from the unique objects that fill the space and locations of sound sources - I do not want to reproduce all of this linearly, it would not be a true representation of my desired sound experience.
So, I am saying that FLAC is ultra-linear, but that is not the way I want my sound experience anyway - its too artificial. Encoding to high rate MP3 may drop some information, but if this information is not a true representation of that original sound, then you are losing information that came with a high degree of uncertainty anyway.
Basically, maybe losing some ultralinear properties could place you in a more approximate location in the room for acoustic recordings, which maybe considered a more accurate representation of the original sound.
Its true that if you have a few mics set up and you use math to convert to the frequency without compression, you could wear a VR headphone headset that reproduces a linear signal from the carefully combined frequencies of all sources. You can't do this without a frequency domain interpretation, since it is arguably a more representative format of sound.
The linear interpretation is the aberration to an acoustic recording, unless you intend that no objects in the room move at all.
So, I quite like the interpretation high rate MP3 gives to some recordings, and would consider it a more accurate representation of the original environment given spatial uncertainty maybe desirable.
As for electronic recordings and artificially created stereo spaces, well, its just how I feel about it that matters, not how linear the signal is.
High rate mp3s - the frequencies you can possibly hear are all there, its not as if just because the file size is so much smaller, its somehow less true - as if making the linear filesizes bigger would continually gain more information about the original source.
end rant - give me full spectrum sound with accurate voicing and some spatial sense with a little uncertainty to it and I'm there!
