[Flac] Re: Odd number of samples in a stereo wave file

[Brian Willoughby]

On Nov 16, 2007, at 09:48, Free Lunch wrote:
> One missing byte is a huge burden and nonsensical?

All audio file formats, including both WAV and FLAC, require complete  
frames containing one sample per channel.  I don't see how you can  
get everyone to agree on whether to drop the sample that's paired  
with the missing byte or add a sample that does not exist in the  
original.  Either way, you don't have any way to losslessly recreate  
a WAV that is badly formatted from the start.

> People post on this list looking for solutions. They don't want to
> become experts in the WAV format (including the undocumented parts).
> They just want to compress their audio without losing any of the
> original (you know, LOSSLESS).  And if some of their original isn't
> included in the archive, they want an exit code to indicate a problem
> and not that everything is okay.

If the original is badly formed according to all standards, then  
there is no way to compress the original without altering it.   
Whether you actually alter the file in a separate step by repairing  
the WAV, or just make FLAC generate a valid compressed file from an  
invalid WAV, the end result is the same: the original data is  
altered.  What you want cannot be described as lossless.  It seems  
like you want repair and compress combined in one magical step.

By the way, in the example that started this entire thread, there is  
an exit code indicating a problem.  Perhaps you are using a very old  
version of flac, because there is no way to compress an invalid WAV  
and think that everything is okay.  The error message could be made  
more clear, but there is an error message.

> Imagine if Sony, Pioneer and other hardware player manufacturers were
> so quick to reject "nonsensical" audio.  Can you imagine if your car
> CD player was so finicky?  How do you think real world customers would
> react?  They don't want to hear excuses about format, they just want a
> product to work. Sorry if that "just work" expectation is too
> "punitive" for you.

I have never seen a CD player from Sony, Pioneer, or any other  
manufacturer that will play a CD which does not follow the  
specification correctly.  Give me an example of someone selling CD  
media which does not comply with the CDDA specification AND that  
still plays in any manufacturer's CD player, and I'll be happy to  
have learned something new.

You're not being very consistent in your demands.  On the one hand,  
you want flawless and reliable lossless compression, but on the other  
hand you want errors to be silently altered so that users are not  
bothered with the problem of choosing which way to repair the error.   
You seem to be missing the fact that there is no way to repair the  
bad WAV data without adding or removing some of the original.

> As I have stated before on this list, that is a completely unrealistic
> fantasy.  I have yet to find a customer who favorably responds to
> being told to change their tools. Dumping the uncooperative vendor is
> generally much easier.  Customers don't want excuses, they want
> solutions.

You seem to have a strong sense of business and what customers want.   
Why are you on this list expecting someone else to solve your  
customers' problems for you?  You have been offered many valid  
options that would meet your customers' needs, but they all require  
you to earn your customers' money by working out the solution for  
yourself.  I don't see how you will meet with success by adopting an  
attitude that someone else should build your business solutions for  
you.  It's not clear to me what you expect, when your customers are  
paying you and yet you are here expecting someone else to solve your  
problems.

> And as pointed out before, sometimes recordings get interrupted.  If a
> recording device loses power it may not be able to write an even
> number of bytes or update the header size. Sending bug reports to the
> manufacturer won't help.  Nor will they help in cases where the
> hardware is no longer being developed.

WAV is the wrong choice for recording format when you expect that  
power might be interrupted.  There are other formats for storing  
audio which are fault-tolerant and more robust when there is a  
potential for power loss.  If you are using poorly-designed recording  
hardware, then you will have poor results.  I don't see how you can  
get around that.  Besides, there is no lossless solution for the  
problem of incomplete recordings.  By definition, a recording that is  
interrupted by a power failure has already suffered from a very  
serious loss of data.

> In the real world, not all WAVs are perfect.  That will never change.

So fix them!

> For many of us who produce terabytes of audio masters, modifying them
> is not an option.  The potential for introducing flaws is too great.
> Simply stating "reject them" is not a solution.

I have no idea why you are collecting terabytes of WAV files which  
have errors.  If the WAV files are not compliant with the  
specification, then why is it so important to preserve them  
losslessly?  For that matter, how can you even be concerned about  
introducing flaws in a WAV files that are already known to be  
invalid?  There is no way for a lossless compression scheme to accept  
a badly-formed WAV file without rejecting it.  Some human input is  
required to either toss the odd sample or generate a sample out of  
thin air to pair with it.

Brian Willoughby
Sound Consulting