| Newsgroups |
gmane.comp.audio.eac.user |
| Message-ID |
<[email protected]> |
On 10/1/20 9:46 AM, Fred Maxwell [email protected] [EAC] wrote:
> {I’m assuming that you have at least two copies of your music
> collection and that you have diff}
Three local (U.S.) and two offsite (France, Norway). I assume I have
diff (I run both Linux and W*nd*ws) but in my head a backed up corrupt
file would still test OK with the diff method. That is, the two files
could be the same as each other but still wrong.
> How many bit flips within FLAC files has your code detected?
I have found a total of two since, I think, 2006 in my own library that
had somehow been corrupted. What I don't know is if they always were and
I never noticed or if something happened. In both cases, the issues were
clearly audible once I knew to listen for trouble. I'm pretty sure that
one of the problems was caused by the then-ancient NAS I was using, the
other one is a complete mystery. Fortunately I was able to easily
replace both files. I have to plug HDTracks: One of the bad files was
one I'd purchased from them and they made a replacement available.
But just last week I downloaded an album and did |flac -t| before going
any further and found that four of the tracks wouldn't validate.
> If you’re seeing bit flips in files, whether FLAC, executable,
> OS, or anything else, you have issues with your storage subsystem
> that need to be addressed.
My primary music drive is now a FreeNAS server using zfs, so I don't
anticipate any more surprises going forward. When I run a FLAC check
(which takes days!) I usually run against one of the backups rather than
the primary. I'd rather find a problem at the end of the chain and work
my way back up, if that makes any sense. It's probably not rational, but
it certainly isn't hurting anything.
I can see the appeal of things like whole-directory MD5sums and the
like, but I'm unlikely to remember to properly recalculate those files
after making tag changes. While I don't make changes every day, I do it
often enough that something like that wouldn't work for me. Having
something like FLAC's checksum that /only/ tests the actual audio is
perfect for me.
> My music (primarily in ALAC) is stored on a eight drive RAID-6
> NAS that performs regular scrubbing (parity checking). I have
> confidence in the data integrity of all files stored on that NAS,
> including the music files. Of course, that is not my only copy of
> my music (or any other files).
I don't think either of us has a whole lot to worry about. But my inner
obsessive nerd just really likes being able to do a real-time corruption
test even though in the real world it's probably very nearly pointless.