I want to write a presentation of different lossless encoders, their advantages, features, disadvantages, risks... Nothing scientifical, just for the common user. I'm familiar with lossy/lossless encoding, and thanks to Speek and Hans Heijden, I have very precise data about speed and ratio. But some questions remains...


First, multichannel :
Are encoders ready for mutichannel encoding (5.1, 7.1...) ? What encoder support this feature, not really useful now, but probably later ? MLP is exclude. I know that's planned for OptimFrog, and that WMA9 lossless is ready. But that's all....

Then, bit depth and frequency resolution :
I generated a high-definition file (24/96), and monkey, flac, wavepack & ofr were able to compress it. LA is 16 bit only for the moment. I couldn't compress with Shorten, and didn't try with Rkau, Lpac & Bonk. Any comment ?

Finally, error recovery or robustness :
The most important thing, in my mind. Are some format more secure than others ? I often read that FLAC is robust, and therefore can read a corrupted file (just by skipping the corrupted part). Maybe this feature can explain the low encoding ratio of Flac format. But what about monkey, wavepack, ofr, shorten and the others ? According to Ghido's agenda, "recovery information at creation and repair for corrupted files" is planned for this format (source). Are others format non-secure, less-secure, for archiving ? I exclude PAR2 and other external solution for eror recovering. Will the slightest error within ape or wavepack file completely and systematically ruin the whole stream ?


In my opinion, wavepack, flac, optimfrog and monkey are the most interesting file format of the moment. I really need the complementary requested answers for these formats to complete my guide. But answers about old format, or newer as LA or Bonk, are welcomed too.
Thank you very much for answers

So far, multichannel and robustness are only features of FLAC.

David Bryant is working on WavPack4, that will add both features (and many more) to the format, and better tuning for the already popular lossy/hybrid mode

There are several components to robustness:

1. At the lowest level is error detection, i.e. the codec can tell when it there's an error in the data and won't pump out junk on the other end.

2. Next is error recovery, being able to restart correct decoding after an error occurs.

3. Above that is error correction, which can actually use redundant information to correct errors.

#1 and #2 rightly belong in the transport layer of a codec, and is what you see in Ogg and native FLAC. #3 could be considered a transport function, or addressed at a higher layer (e.g. adding redundancy with PAR2).

Josh