Install zsync with Homebrew, apt, dnf, MacPorts, Nix, pacman, zypper

Project links

• Homepage
• Repository
• Package manager page

Commands and aliases

• zsync
• zsyncmake

Project history

Colin Phipps developed zsync to solve a distribution problem rather than a private synchronization problem. The original site contrasts it with rsync: rsync is designed for synchronizing data between machines, while zsync is aimed at distributing one file from a server to many downloaders without running a special server daemon.

The project was originally released in 2004. The technical paper describes the core idea as client-side rsync: the server performs one-off checksum precomputation into a control file, while each client uses HTTP range requests to fetch only missing content.

The 0.7 line is a rewrite of the program. The current site notes that it is less mature than the 0.6.x line, removed older look-inside-compressed-file support, and keeps development history in the GitHub repository.

Adoption history

zsync's best-known adoption niche is large-file distribution where many users already have yesterday's or last week's file: Linux installation images, daily builds, virtual-machine images, and other release artifacts. The server page explicitly calls out daily Linux installer or live images as a good fit.

The project never became a general directory-sync replacement, and its own documentation says multiple-file support is not planned. That boundary is part of its packaging identity: zsync is a small CLI for one-file HTTP distribution, with zsyncmake producing the metadata publishers need.

How it is used

Publishers run zsyncmake against a target file to generate a .zsync control file containing the precomputed checksums and one or more URLs for the full content. The web server must make both the .zsync metadata and the full target available, and should support HTTP Range requests.

Users run zsync against the .zsync URL. The client scans existing local data, downloads only missing ranges from the target URL, and reconstructs the new file. The design is most effective when old and new files share content and when compression is piecewise or rsync-friendly.

Why package nerds care

zsync is package-nerd significant because it attacks bandwidth costs at the distribution layer using boring web hosting. It asks maintainers to publish one extra metadata file, not to operate rsync infrastructure or invent a new protocol.

It is also a good example of a package whose constraints matter as much as its feature list: single-file distribution, HTTP range behavior, stable target content, and compression choices all determine whether the package is magic or useless.

Timeline

• 2004: zsync originally released.
• 2000s: 0.0.x through 0.6.x releases accumulated on the official download page.
• 0.7.0: Rewrite line introduced, with the project warning that it may be less mature than 0.6.x.
• 0.7.2: Current version listed on the official download page.

Related projects

• Related projects and technologies include rsync, rdiff, HTTP/1.1 range requests, gzip --rsyncable, Linux distribution ISO publishing, update systems, and mirror networks.

Sources

• https://github.com/cph6/zsync
• https://zsync.moria.org.uk/
• https://zsync.moria.org.uk/downloads
• https://zsync.moria.org.uk/paper/
• https://zsync.moria.org.uk/server