This is strictly accurate, but the argument against recompression is easily overstated.

Audio compression systems combine lossless compression with psychoacoustic compression. For present purposes we can ignore the lossless element. The psychoacoustic compression is carried out differently with different compression systems, but is essentially based on the same set of rules derived from our understanding of how the human ear hears. Eg, ignore all data above such and such a frequency. Eliminate data for particular frequencies occurring at the same time as other particular frequencies, because they'll be masked anyway. And so on.

AAC -> WAV/AIFF incurs no loss -- we're simply expanding the compressed data

WAV/AIFF -> MP3 will technically lose data, but (providing the squash isn't too tight) will mostly be throwing away data that isn't there anyway, because it's operating by much the same filter/masking rules as the original AAC compression.



---
el bid