File Hosting & Sharing

Instant Rollback

Definition

Where file versioning keeps the history, rollback is what spends it. The earlier bytes are already sitting in storage, so restoring them is a pointer change — mark the previous version current — rather than a transfer, and the time it takes is measured in milliseconds regardless of whether the site is 2 MB or 2 GB. The operation moves forward rather than backward: the restore is itself recorded as a new version, so you can roll forward again to what you just left, and the history never loses an entry. Done properly it is atomic across the whole set of files, so visitors never see a page from the new version paired with a stylesheet from the old one.

Why It Matters

The number that matters is how long a broken page stays broken. Re-finding the previous build, re-zipping it and re-uploading 200 MB over a domestic connection is comfortably fifteen minutes, and it assumes the old build still exists somewhere, which is exactly the assumption that fails under pressure. A pointer flip is under a second. What remains is the edge: if the responses carry Cache-Control with max-age=3600, copies already cached can keep serving the broken version for up to an hour, so a rollback that does not also trigger cache invalidation only half works.

How It Works

Pick a version from the history and the host copies that version's identifier into the current position, usually by writing it as a new version so the record stays append-only. All files in the set switch together, which is what makes the change atomic rather than a rolling half-state. The host then purges the affected paths at the edge so the next request reaches the origin instead of a cached copy, and short-lived HTML paired with long-lived hashed assets keeps that purge small. What rollback does not touch is everything outside the files: DNS changes stay changed, form submissions already captured stay captured, and an external database is entirely unaffected.

Real-World Example

A team pushes a new hero image to nova-launch.99helpers.site an hour before a product announcement, and it overflows the layout on every phone. Rather than debugging live, they open the version list, select the copy from that morning and restore it. The site is correct again in a few seconds, the edge cache for the HTML is purged as part of the same action, and the broken version stays in the history for someone to fix properly on Monday.

Common Mistakes

  • Rolling the files back but leaving a long max-age in place, so returning visitors keep seeing the broken version from cache
  • Expecting a rollback to undo things that were never files — captured leads, analytics events and DNS records all stay as they are
  • Leaving versions unlabelled, so the history is a column of near-identical timestamps and nobody can say which one was last known good
  • Restoring in a panic to whichever version came before, without checking that it was ever working — two bad releases in a row make that a coin flip

Related Terms

Put a file online in seconds

Drop in a document, an image, a page or a whole static website and share the link — free, with no build step and no server to set up.

Host a file free →