If you are like tk3, say if you have the HDD space and the shortened battery life isn't an issue, I'm all for lossless. I even have some FLAC files on my player.
However, I would also argue that Mp3s with a high-enough bitrate are as good as you are going to need on a portable player. The better the bitrate, the more accurate the algorithm can be when encoding each sample. The algorithms which generate mp3s do remove stuff from the music, but they are designed to remove the stuff you won't hear anyways. Granted they are not perfect, but they cut back on stuff like:
really high and really low frequencies that are out of most people's audible range. And if they aren't out of your range, they are so damn close that you wouldn't miss them. raise your hand if you can still hear 20k... put your hand down, you aren't fooling anyone. You wont even hear 2k on lossless when a police car drives by with its sirens on.
masked frequencies. tones you cannot hear anyways are cut out. If you wouldn't have heard it, why not cut it? (a masked frequency is a quieter tone at a close frequency to and at the same time as a louder tone. our brain does not process these).
bass stereo channeling. our brain has a really tough time localizing low frequencies. since we cannot tell where the bass is coming from (and most of the time it's just front and center), the information which encodes the stereo panning for bass is cut.
sounds below the threshold of hearing are not just cut, they are removed entirely. again, if you weren't gonna hear them before, who cares?
Also, and this is less relevant, but just an fyi, since most people don't know, but the first step in the mp3 conversion process is the entire lossless encoding process. So it undergoes huffman coding and entropy coding first, along with a few other procedures. So you see, a lot of the file reduction doesn't remove any of the music at all!
Furthermore, bit resevoirs allow the algorithm to only use the number of bits it needs to encode each sample (this applies even to constant bit rate mp3s, vbr is even more efficient at this). So the simpler a sample is, the less memory it will use up, and those unused bits are still available for when the music gets complicated again (this saves you space twice). In case you were wondering, it uses just enough bits per sample so that the quantization noise will be inaudible or maskable (The generated noise can then be "shaped" so that it is hidden underneath the music and inperceivable. remember what i said about masked frequencies? They can be used to our advantage too!). This greatly reduces file size without sacrificing anything from the sound, but isn't done on lossless because it technically is loss.
Another thing that people don't realize is that not all algorithms are created equally. There are different ones that can produce the same song at the same bitrate with differing sonic qualities. Sure, all of them try to do the same thing, but since they use different methods, the outcome is different. This is usually very slight however. The LAME encoder is supposed to be the "best" in terms of quality in that it does the best job of keeping dynamic range up and preserving audible spectra.
I know all this crap isn't going to change anyone's mind, but I hope it helps those on the fence. In the end, you do sacrifice some of the information that was intended for you to hear, but most of it you wouldn't have heard. so for those saying "Oh my god you spent so much on your portable rig and you are gonna use lossy?!" I feel fine saying, "yes" cause I can listen to more tunes than you since my battery will last longer, and when I listen to portable, the noise from the environment trumps the noise and loss from the mp3 algorithm by several orders of magnitude even with my closed-air ATH-ESW9. At home when it's quiet, I want CD quality or better, out and about, what's the point? (I should mention that I live in Manhattan, where it is never quiet. This argument wouldn't apply to those who are lucky enough to be able to listen to their portable audio in peace).
