Powered by OPF
OPF Q&A Static Archive
133 questions · 234 answers · 193 comments · 172 tags · 118 contributors · archived August 2026
Archived copy. This is a static snapshot of the OPF community Q&A (qanda.digipres.org), preserved from its Question2Answer database. It is read-only and nothing here has been edited since capture in August 2026. To ask new questions, use the OPF community on GitHub.

Parity data for ISO images: anyone doing this? Best practices?

1
votes

Earlier today I had a discussion with a colleague on possible ways to store (ISO) images of CD-ROMs and DVDs in our repository system. In addition to checksums, he suggested to also generate and store parity data for each image file (e.g. using the par2 tool: http://manpages.ubuntu.com/manpages/trusty/man1/par2.1.html). This would enable one to repair files in case of bit-level corruption.

This made me wonder how commonly parity info is used in digital archives, and if there are any best (or at least recommended) practices. E.g. what levels of redundancy are typically used? I'm not really familiar with this at all, and it's not a subject that is often mentioned in digital preservation discussions (but see https://twitter.com/anjacks0n/status/733281959762395139).

  • I should probably have said forward error-correction codes or erasure codes, as parity codes are just one tactic. This kind of thing is usually delegated to the storage sytem, but having explicit parity files may be a better idea as they are portable across storage platforms.
    — anjackson, 25 May 2016
  • Thanks Andy. Your remark about the storage system already taking care of this is something that crossed my mind as well, and made me wonder whether repeating this using parity files wouldn't be overkill.
    — johanvanderknijff, 25 May 2016

1 answer

1
votes
✔ accepted
Thought I should try to give an actual answer here and not leave it for more years.

I don't know of anyone in the digital preservation community that is doing this and I suspect Andy's comment speaks to why. So I also don't know of any best practices. It's not the same as saying that absolutely nobody is but it seems likely that nobody is so hopefully answers your question.