Preservation
The question you ask too late
We almost always discover the true state of a backup at the worst moment: the one when we actually need it. Between “the file is there” and “the file can be recovered” lies a gap that only a tested restore can measure. A backup you’ve never restored isn’t a backup, it’s a hypothesis.
What breaks a restore
An archive can be present, the right size, and still unusable. The classic causes:
- Corrupted archive: a single byte altered by a bad sector or an incomplete transfer, and extraction fails or produces unreadable files.
- Lost key or password: an encrypted archive without its key is mathematically unrecoverable. Encryption protects against you, too.
- Incomplete structure: a missing volume, part of a split set, or linked files (sidecars, derivatives) left elsewhere.
- Missing manifest: without a reference inventory, you can’t know what should be there, you can’t prove nothing is missing.
None of these failures show up when you glance at a folder’s size.
Testing means actually restoring
A useful test doesn’t stop at “let’s see if the file opens.” It means restoring a sample to a separate location, then verifying:
- that the archive extracts without errors;
- that the checksums of restored files match the manifest;
- that the files actually open in real software;
- that the decryption key is accessible and works;
- that associated metadata and sidecars are present.
A representative sample (a few RAW files, one derivative, a file with a sidecar) is enough to reveal most chain problems.
How often?
There’s no magic number, but a few reasonable markers:
- With every important new archive: an immediate sample test.
- After any change of media, off-site destination, or tool.
- Periodically for dormant copies (once or twice a year), because media degrade in silence.
The goal isn’t paranoia, it’s turning restore into a routine gesture rather than an improvised emergency.
In practice
- Never judge a backup on its presence alone: judge it on a restore.
- Confirm today that you know where your decryption keys are.
- Restore a sample to a test folder, not over your originals.
- Record the date of your last successful test, the way you’d log a service check.
Obscura Flow treats restore as a normal step in the workflow: it verifies the signature before extracting, compares against the manifest checksums, and handles conflicts without accidental overwrites. As a natural extension of verifiable archives, testing a restore becomes a reflex, not a gamble.