Taking the backup back into your own hands

Cloud backup services start to feel uneasy after a few years, and the reason is structural rather than technical: the data sits on somebody else's server, and one missed payment or one closed service can take years of backups with it. This project arranges things the other way round. The software is free and open source, and the storage is whatever you already have — object storage, a WebDAV endpoint, an FTP server, even a disk on the same desk. The tool does the encryption, the chunking and the incremental upload; the bill and the ownership stay with you.

Encrypted before anything leaves

Encryption happens on the machine, before a single byte goes out, with AES-256. What the cloud holds is ciphertext and nothing else, which means the storage provider cannot read the contents even if it wanted to. Jobs are configured in a browser interface rather than a desktop dialog: filters for what to include, schedules for when, bandwidth limits for how much of the line to take, incremental runs for the daily case and a full pass periodically.

The 2.4.0.0 stable release, out on 3 September 2026, is the largest in years. Whole-disk imaging used to be a Windows-only affair and now covers Linux and macOS too, so a raw backup and restore can be taken per disk or even per partition. A new sync mode adds lightweight one-way file copying, which is a different job from a versioned encrypted backup and is now honest about being one.

Fighting bit rot

Cold backups rarely die of deletion. They die of silence: data sits in the cloud, quietly rots, and the discovery happens at the worst possible moment — during a restore, when there is no time to negotiate. The PAR2 verification introduced here generates recovery data and uploads it alongside the backup volume, so damage on the storage side can be repaired rather than mourned.

Alongside it come live progress reporting, desktop notifications on every supported platform, and cloud backup for an office suite. None of those are dramatic features, and together they are what moves a long-running open-source project from interesting to dependable.

What it is not deserves the same clarity. It is not a two-way sync client, since the new mode copies in one direction only. It is not a hosted service, so there is no support line to call at three in the morning. And it does not hold the key for you: the passphrase is the whole security model, and a lost passphrase means a lost backup, forever, with no recovery path and nobody to appeal to. That is the price of the design, and it is a fair one.

The 3-2-1 rule, and the part people skip

Everyone repeats the rule: three copies, two media, one of them elsewhere. The first two are usually done, because they are local and cheap. The third is where people stop, and the reason is not laziness but procurement — an offsite copy means a provider, a contract, a price that creeps upward and a nagging worry about what happens if the provider changes hands.

Separating the software from the storage removes most of that objection. The offsite leg can be an object storage bucket, a friend's server behind a WebDAV endpoint, or a disk in another building, and switching between them is a configuration change rather than a migration. A schedule that runs daily, 365 days a year, without anyone remembering to plug anything in, is not glamorous; it is simply the thing that works.

What it looks like after six months

The honest picture of a backup set-up is not the first evening but the two hundredth day. By then the incremental runs take minutes rather than hours, because only the changed blocks travel; the storage bill has stayed small because nothing was uploaded twice; and the only interaction required has been reading a notification that said the last job succeeded.

That is the point at which a backup becomes real rather than aspirational. A system that demands attention every week gets abandoned in a busy month, and an abandoned backup is worse than none at all, because it produces confidence without coverage. Setting the schedule once, letting it run daily and checking the log when the notification looks wrong is the whole discipline.

Two habits that go with it

The first is a restore test. A backup that has never been restored is a hypothesis, and the test costs ten minutes: pick one folder, restore it somewhere else, compare. The second is keeping the passphrase somewhere that survives the machine it protects — a printed copy in a drawer is unglamorous and works, while a note on the same disk is a joke with a delayed punchline.

In the 3-2-1 rule, the offsite leg is the one most easily captured by a cloud provider. Encrypting with AES-256 and choosing your own storage hands the choice back: free software, a backend you can replace, a key you hold yourself. The long life of data has always belonged in the hands of whoever owns it.