Hi Tom,
> On May 18, 2015, at 12:44 PM, Tom Fine <[log in to unmask]> wrote:
>
> Hi Dave:
>
> Can you put some of this in English?

Gladly.

> What is "demux" and "mux"? It sounds nasty but I assume it's an abbreviation for something?

Sorry these are short for multiplex and demultiplex. These terms are often paired with encode and decode, Muxing has to do with the container or file format whereas decoding is about the stream contained within.

So for instance you may have a WAV file but would have to demux it to know what the contents are. Then depending on what the content is you have be able to decode it (it's a two part step). Sometimes you may have a file that you can demux but has contents that you can't decode (like when Quicktime opens a player window at the right size and duration but then can't play the content). This is usually a behind the scenes assembly line.

Reto Kromer recently did a nice presentation on this, see slides 38-42 at http://www.reto.ch/training/2015/20150424/20150424.pdf.

> Also, what I think you're saying is that FLAC is just another audio CODEC, as is WAV. Both are lossless in that they do not throw off any samples or bits, and neither impart any perceptual encoding. Is that correct?

The language here is a little tricky because FLAC is often used to refer to FLAC the file format (.flac) as well as FLAC the lossless audio encoding. So FLAC is both a stream and a container, similar to MP3 I suppose. While a FLAC file will contain a FLAC encoding, the FLAC encoding may be in other file formats such as QuickTime and even WAV.

> Here's what I don't get about this discussion. Why would someone want to use FLAC as their primary lossless format rather than WAV, given that FLAC includes (and features) a data-compression scheme that inherently puts more data at risk per potentially bad sector in a storage device? Given the low cost of storage, why not opt for less data at risk per sector?

Cost is really a secondary issue compared to the fixity advantages.

This density issue is true. A flipped bit on a FLAC will have more impact than a flipped bit in PCM/WAV. On the other hand a FLAC file will use about 3 times less bandwidth, be 3 times faster to fully read or checksum, and faster to move from one disk to another. I'd suggest that moving 30 TB of data to new media has a slightly greater associated data risk than moving 10.

But the reason to use FLAC is really the next part. A flipped bit in PCM/WAV causes one sample somewhere to be different (maybe enough to make a click or it may be very hard to discern). If the archivist kept an external checksum, they may notice that the PCM/WAV file has a checksum mismatch but the PCM/WAV file makes it nearly impossible to figure out what the impact is. If the archivist has no checksum then the damage to a PCM/WAV is hard to be aware of.

A flipped bit on a FLAC will firstly create a mismatch during decoding. So reading a FLAC file should warn that the file is damaged and that the contents no longer represent what was originally encoded. Furthermore because audio frames within FLAC contain CRCs the damage is noted down to a portion of a section. I've worked with damaged FLAC files before and it is very easy to known that there is damage and pinpoint where. Because of the CRCs audio players can use error concealment (similar with the error concealment in a DAT player) to reduce the extent to which the error is noticed. I wrote a lot more about FLAC's fixity and preservation features here: http://dericed.com/papers/reconsidering-the-checksum-for-audiovisual-preservation/.

But in general I'd suggest that PCM/WAV is a more vulnerable format than FLAC. It requires a lot of external efforts to ensure that the data of a PCM/WAV is unchanged whereas these concerns are built into FLAC as a required feature. To have a FLAC simply by itself provides a means to validate that the data within purely represents the original audio that was encoded and if not where the discrepancy lies, PCM/WAV has no such features and requires potentially complex external fixity measures to try to accommodate this. With PCM/WAV an error or the need to apply error concealment is programmatically guesswork whereas the CRC and MD5 protection mechanisms of FLAC make this information very clear. Other audio formats use internal CRC protection to validate transmission such as MP3 but FLAC is one of the few preservation formats to make this feature available as well.
Best Regards,
Dave Rice

> -- Tom Fine
>
> ----- Original Message ----- From: "Dave Rice" <[log in to unmask]>
> To: <[log in to unmask]>
> Sent: Monday, May 18, 2015 12:11 PM
> Subject: Re: [ARSCLIST] Is it time to rethink FLAC ?
>
>
>> Hi Lou,
>>
>>> On May 18, 2015, at 11:49 AM, Lou Judson <[log in to unmask]> wrote:
>>>
>>> I will admit to not reading the entire post, but have a question:
>>>
>>> According to my understanding, FLAC is a non-lossy compression scene applied to WAV and PCM files,
>>
>> A FLAC encoding isn't limited to PCM/WAV source, the input to the FLAC encoder is simply the raw audio with contextual metadata (bit depth, channel count, sample rate, etc) from another source.
>>
>> For instance you could have
>>
>> WAV ---demux---> PCM ---decode---> rawaudio ---encode---> FLAC codec ---mux---> FLAC file format
>>
>> but you could have
>>
>> MP3 ---demux---> MP3 ---decode---> rawaudio ---encode---> FLAC codec ---mux---> FLAC file format
>> or
>> S/PDIF ---demux---> PCM ---decode---> rawaudio ---encode---> FLAC codec ---mux---> FLAC file format
>>
>> Most of my archival experience with FLAC involved no WAV file at all but the FLAC resulted directly from a digitization (analog source) or data migration (CD-R, DAT, etc).
>>
>> It is just as technically feasible to digitize directly to FLAC as it is to scan documents to lossless LZW in TIFF or digitize video to JPEG2000 or FFV1. It doesn't matter what type of decoder is producing raw audio data but only the the raw audio data is the input to the FLAC encoder. NB I'm taking behind the scenes in the transcoding or recording process, it's rare that an operator would be connecting these pieces together without some form or programming.
>>
>>> not a digital encoding format in itself.
>>
>> I would consider FLAC as a digital encoding in itself.
>>
>>> If that is so, then one must start with WAV (or other PCM format) files in order to get to FLAC. Therefore FLAC is an accessory, not a proper format.
>>
>> I disagree. The raw audio data delivered to a FLAC encoder may come from any source that can provide raw audio data and the necessary metadata. This could be a WAV demuxer and PCM decoder but there is no such limitation that it must be. It could be from various digital audio transmission standards, other decoded lossless or lossy encodings, or other uncompressed audio encodings.
>>
>>> If this is so, then it can only be seen as a storage format, not a recording format, and the argument is academic.
>>
>> * head desk *
>>
>>> Intelligent refutation is welcome.
>>
>> I can refute, but the information above seems to be based on presumptions (correct me if I'm wrong). If you have a citation that claims that "FLAC is not a recording format", that "FLAC is not a digital encoding format", or that "one must start with WAV (or other PCM format) files in order to get to FLAC" than I am happy to refute.
>> Dave Rice
>>
>>> <L>
>>> Lou Judson
>>> Intuitive Audio
>>> 415-883-2689
>>>
>>> On May 18, 2015, at 6:39 AM, Dave Rice <[log in to unmask]> wrote:
>>>
>>>> Hi all,
>>>>
>>>> I was very pleased to hear that the merits of FLAC as a preservation format were again being considered by ARSC List; however most of the discussion only considered a few of its advantages over PCM/WAV such as that the openness of the format and resulting storage requirements, but the thread hasn't yet covered FLAC's preservation and fixity features over PCM/WAV.
>>