It is my understanding that Apple lossless is M4A not mp3 M4a is not a compressed file rather it is a zipped file with full information.
M4A is a file container (MPEG-4) and can include audio encoded with either lossy AAC or lossless ALAC. Typically, iTunes files are encoded with AAC.
2 Likes
It depends on where the boundaries are for consideration 
Starting with 96k/24 -> [ 44.1K/16 -> CD encode/write -> CD -> CD read -> 44.1K/16 ] -> DAC - the bit in [] is lossless ignoring CD read errors.
Starting with 96k/24 -> [ 44.1K/16 -> FLAC encode -> Stream/File -> Flac decode -> 44.1K/16 ] -> DAC - the bit in [] is lossless ignoring transport/file errors which are probably less likely than CD read errors.
Starting with 96k/24 -> [ 44.1K/16 -> MP3 Encode -> Stream/File -> MP3 decode -> 44.1K/>=16 ] -> DAC - the bit in [] is lossy because the output of mp3 is not bit identical to the input to mp3 encoding.
The argument that all end to end chains are lossy is part of the propositions behind MQA and arguments for it as they are trying to achieve something close to an analog end to end lossless process - there are more than enough thread arguing that point, so no need to go into that here
On the subject of why 44.1k/16 -> mp3 -> 44.1/24 is ok…
This is because within mp3 decoding there are frequency domain samples (obtained from decompress of the stream blocks) which are handled as floating point and subject to high precision calculations to convert the frequency domain samples into time domain samples before eventually being converted to integer and dithered to 44.1k/16 or 44.1k/24.
As truncation and dithering will always introduces additional noise due to quantization, then it follows that quantizing to smaller intervals will yield less additional noise from mp3, so decoding mp3 to 24 bit instead of 16 bit makes some sense and would tend to place the noise introduced by decoding down below the noise floor of most playback equipment whereas 16 bit decoding probably will not.
Maybe this doesn’t really matter as one may argue that noise introduced by encoding vastly exceeds this anyway, OTOH, why add more noise when you can easily avoid it?
Yup - I understood all of that… (Grams: cue sound effect of nose elongating)
1 Like
I’m guessing first bit made sense re lossless / lossy 
Re: time and frequency domain samples - if you have seen a waveform graph such as Roon sometimes displays for a track, then that is time domain samples - ie signal level samples are taken at points in time - this is what we are all used to with PCM, and talk of sample rates etc.
If you have seen the display of a spectrum analyser that shows the frequency composition of an audio signal with various bars indicting the signal intensity of various frequency components, then that is basically a frequency domain view of audio (except that there are vastly more frequency bars).
If you were to take rapid photos of the of the display, then that is kind of like a sequence of blocks that represent frequency domain samples. That is some complex math to convert from one to the other. The advantage of processing audio in the frequency domain is it is easy to look for frequency bands with very little information in them and basically just bin them on the assumption a listener wont notice 
That is very (and overly) simplistically how mp3 encoding works. Decoding has to do the opposite (but by now some frequencies have been binned form some blocks so are not recoverable - hence it is lossy). There is of course a lot more to the process particularly on the encoding side to decide what to throw away.
(Caveat - my memory of this is very fuzzy these days, so maybe someone more up to date and who know this better may chime in and correct me…)
1 Like
Over the years there was always confusion to the difference between lossless and lossy. These two words are too similar in pronunciation.
Should keep the word “lossless” because it gets to the heart of the meaning but instead of “lossy” the word that should be used is “degraded” even though for some this rip sounds great, in the end it has been degraded.
–MD