Duplicity

Ben Escoto created Duplicity in 2002 to combine librsync delta transmission with standard GnuPG encryption for untrusted remote storage. Decades of maintenance have turned it into a dependable Unix utility for automated offsite backups across dozens of storage protocols. That heritage means it treats destinations as passive targets, keeping encryption keys and scheduling strictly under the administrator's local control.

Why Duplicity instead of Carbonite

Carbonite provides a managed backup service centered on a proprietary desktop client and bundled cloud storage, charging recurring fees per endpoint. In contrast, Duplicity emphasizes modular integrations and open protocols. Rather than locking backups into a single vendor's vault, Duplicity connects natively to SFTP, SSH, rsync, and WebDAV, or routes archives to object storage through an integrated Rclone backend. Teams can send encrypted volumes to Backblaze B2, AWS S3, or self-hosted MinIO without installing proprietary daemons.

The operational distinction lies in automation and control. Carbonite automates continuous desktop sync via a graphical interface, while Duplicity plugs into cron jobs, shell pipelines, and systemd timers. You trade Carbonite's managed portal and instant GUI restores for protocol independence, reproducible Unix tooling, and zero software licensing costs. Administrators manage their own GnuPG keys and define their own storage lifecycles without vendor lock-in.

What it does well

• Broad storage integration: Connects natively to SFTP, rsync, WebDAV, AWS S3, and Azure Blob, plus dozens of cloud backends via Rclone.
• Bandwidth-efficient transfers: Uses librsync to upload only changed byte ranges within modified files during incremental runs.
• Standard GnuPG encryption: Encrypts and signs archives locally before transmission, preventing remote hosts from inspecting data.
• Point-in-time recovery: Restores directory trees as they existed on any specified date for predictable rollback drills.
• Unix metadata preservation: Retains user permissions, numeric IDs, symlinks, and timestamps across backup cycles.
• Headless automation: Operates cleanly via CLI flags and environment variables inside scripts and container entrypoints.

Limitations

• Slow sequential restores on long chains: Restoring individual files requires processing the full backup and each successive incremental delta volume in sequence.