I am wondering about this topic, maybe a silly question but anyway
here it goes.
1. Am not 100% sure how a encoder deal with the input,
what i had in mind was that a encoder did take a "fingerprint/snapshot"
of the source, and calculated the compression based uppon that.
I am guessing that this prosses is a fact, but in case it is so,
the decoded file should be a mirror of the input.
This is not the case, a encoder can couse frequency cutoffs and
mess upp the dynamic range, from realy bad to bearly audible.
depending on preset/command line and encoder.

So what i realy wonder about, is if there are any thoughts/ideas
of adapting the wavepack correction to the lossy encoders.?
In the mid 90`s, there was a group, managing to compress
a cd image to floppy sise. (?unbeliveable?)
so am thinking there would be a way to compress audio files
+ correction info, in that manner to.

As i said, it is maybe silly, but in theory it should then boost
the overall quality. And also make lossy compression safe to use
regarding exact quality of the inn/out.
it do mean bigger files, but personaly i think most people is more oriented
to quality then quantety.

i think you're referring to 2-pass compression - the first pass determines the bitrate needed for the 2nd pass, and the best places to allocate bits.



that sounds a little farfetched, but if you could provide a link i'll be very interested.

certainly if a CD image could be shrunk _losslessly_ to a floppy size, then the RIAA would be in deep, deep trouble. as it appears, this has not happened yet (it's difficult to get even 1 song onto a floppy).

lossless compression doesn't often go below 50% of the original file size.

No can do, i can not give a link right now, couse i don`t remember it`s name.
Beside it`s "WZ". (by the way, it was a psx bootimage for PS1.)

About best places to allocate bits, a two pass prosses will boost the quality indeed,
but that`s not exactly what i mean. (i think.) What i had in mind was a way for the given encoder, to read & copy the source info, (levels, headroom etc.)
and keep it, then encode based uppon that information.
When decoding, the information should give a identical output from input.
I asuming this is what goes on when using a lossless engine.
What is realy the practical difference between lossy and lossless.?

If I'm not mistaken, when you make an iso-image of a CD, it will take up lots of space, because it makes an image of everything, even the empty space. If the CD only contains like 2mb of data, such an image would highly compressable, since it would be 2mb of data and like 600mb of 0's. Compressing that to fit a floppy wouldn't be impossible.

hmm.. so you mean a "description" of sorts is determined in the first pass?

i'm thinking the kind of description that leads to an encoder producing lossless output would be so large as to be impractical. that old problem of describing complex systems - that the simplest description would be the system itself. (okay, that's a bit esoteric, but...)

the best you can do is make the output sound the same as the input, but concessions must be made if you're wanting to reduce the file size. a musical waveform is just too complicated and unpredictable.

of course, a MIDI CD would fit on a floppy disc and keep the actual music (sans voice).