What Is an RZ File? (The Long-Distance Compressor Built for Huge, Redundant Files)

What an RZ File Is and Who Built It

An .rz file is the output of rzip, a compression program originally written by Andrew Tridgell — the same programmer behind Samba and rsync — as part of his PhD research, with Paul Russell later contributing substantially to a version 2 rewrite. The latest stable release, version 2.1, dates to February 2006, and the project has stayed largely unchanged since, which is a deliberate reflection of what rzip was built to do rather than a sign of neglect: it solves one specific compression problem and hasn't needed to chase new ones.

That one problem is long-distance redundancy — data that repeats itself, but with hundreds of megabytes of unrelated content sitting in between the repeated parts. Ordinary compressors like gzip use a sliding window of only 32 KB, so two identical 10 MB chunks sitting 500 MB apart in the same file look, to gzip, like two completely unrelated stretches of data; rzip's own project page states plainly that "if you have ram to burn and want the best possible compression rate then rzip is probably for you," aimed squarely at large backup files, log archives, and disk images where this kind of far-apart repetition is common.

Rzip is distributed as command-line source code rather than a GUI application, and its manual page documents ten selectable compression levels, from -0 (fastest, weakest compression) through -9 (slowest, strongest), with -6 the default when no level is specified — a scale that maps directly onto how much of that large matching window the program is willing to spend time and memory searching through on a given run.

How Rzip's Two-Stage Compression Actually Works

Rzip compresses in two distinct passes. The first pass runs an LZ77-style matching algorithm across a history buffer that can reach up to 900 MB, finding and encoding large chunks of duplicated data no matter how far apart they sit in the file — a scale several orders of magnitude beyond gzip's 32 KB window or even bzip2's 900 KB block size. The second pass takes that already-shrunk output and compresses it further using bzip2 itself, applying the Burrows–Wheeler transform and Huffman entropy coding to chunks of roughly 900 KB.

Because the first pass has already removed the long-range duplication before bzip2 ever sees the data, rzip in practice often finishes faster than running bzip2 alone, even though it's doing more total work — bzip2 simply has much less data left to process by the time it gets the pre-shrunk stream. On the project's own benchmark using the Samba source tree, rzip reached a 9.55x compression ratio against bzip2's 4.78x on the same data, a difference the project attributes directly to the long-range matching stage that bzip2 alone doesn't have.

What Using RZ Gains You and What It Costs You

• Gain — dramatically better ratios on redundant large files: the 900 MB window catches repetition that gzip and even plain bzip2 physically cannot see, since their windows are hundreds of times smaller.
• Gain — often faster than bzip2 alone: the first-stage long-range matching reduces the data volume before bzip2 ever runs, cutting its actual workload despite rzip doing an extra pass.
• Lose — no pipelining support: rzip cannot read from standard input or write to standard output, unlike gzip or bzip2, so it needs a real, seekable file to work with rather than fitting into a Unix pipeline.
• Lose — heavy memory use: a typical run on a large file consumes hundreds of megabytes of RAM, a direct cost of maintaining that much larger matching window in memory.
• Lose — no encryption and no multi-file archiving: rzip compresses a single stream, exactly like gzip or bzip2 do, so bundling more than one file first requires a separate tool like tar.
• Lose — near-total obscurity outside its niche: mainstream archive tools like 7-Zip and WinRAR have no built-in .rz support at all, restricting practical use to systems where rzip or a compatible tool is deliberately installed.

Which Programs Actually Open an RZ File Today

The original rzip binary itself, distributed under the GNU GPL v2 and later, remains the most reliable way to both create and open .rz files, and it's packaged in the repositories of several major Linux distributions including Debian and its derivatives. PeaZip, the free graphical archiver for Windows and Linux, lists RZ among the formats it can handle, giving Windows users a path to open rzip output without compiling the original command-line tool themselves.

Mainstream archive managers — 7-Zip, WinRAR — have no documented native support for the .rz extension specifically, which keeps rzip firmly in the category of a specialist Linux/Unix tool rather than a general-purpose archive format anyone can assume their system already handles. Anyone receiving an .rz file on a system without rzip or a compatible tool installed genuinely cannot open it through everyday means, unlike a .zip or .7z that almost any archive manager recognizes on sight.

Rzip's source code has also been packaged as a straightforward build from source on macOS and other Unix-like systems, since the program itself has no platform-specific dependencies beyond a standard C compiler and the bzip2 library it calls into for its second compression stage. There is no dedicated Windows binary maintained by the original rzip project itself, so Windows users depend specifically on third-party tools like PeaZip rather than the original author's own software.

Ubuntu and Debian both still carry rzip in their standard package repositories today, installable with a plain apt install rzip, which keeps the original tool reachable on current Linux systems even though the code underneath hasn't changed in years — a contrast with formats that depend on an actively updated codebase to keep working correctly on newer operating system releases.

Real Problems Documented With Rzip in Practice

The project's own documentation lists rzip's inability to pipeline as a genuine, acknowledged limitation — since it can't read from a pipe or write to standard output, it has to be pointed at an actual file on disk, which breaks the common Unix pattern of chaining commands together with pipes that gzip and bzip2 both support natively.

A second documented issue is memory consumption on very large inputs — maintaining a matching window that can span up to 900 MB means the compressor itself needs a correspondingly large amount of RAM to hold that window while scanning, which the project's own documentation flags as a real, practical constraint on machines with limited memory, not a rare edge case.

A third real limitation is that rzip's development effectively stopped at version 2.1 in 2006, and its successor project, lrzip (Long Range Zip), uses an entirely incompatible file format — meaning files compressed with the newer, actively maintained lrzip cannot be opened with the original rzip tool or vice versa, despite the similar names and shared long-range-matching concept, which has caused real confusion for people assuming the two are interchangeable.

RZ Measured Against Gzip, Bzip2, and Its Own Successor

Common Questions About the RZ Compression Format

What does the .rz extension actually stand for?
It's the output of rzip, a compression program whose name combines "r" for its author's long-range matching approach with "zip" as a generic nod to compression, rather than any relation to PKWARE's ZIP format.

Why would rzip compress a file better than bzip2 on its own?
Because rzip's first stage catches repeated data across a 900 MB window before bzip2 even runs — on the project's own Samba source tree benchmark, that combination reached a 9.55x ratio versus bzip2 alone's 4.78x on the same data.

Can I open an RZ file on Windows?
PeaZip, a free archiver for Windows, lists RZ among its supported formats. Mainstream tools like 7-Zip and WinRAR have no documented native support for it.

Is rzip the same thing as lrzip?
No. Lrzip is a separate, actively maintained successor project using an incompatible file format — files made with one cannot be opened with the other, despite the similar name and shared long-range-matching concept.

Why can't I pipe data directly into rzip the way I can with gzip?
Rzip's own documentation lists this as a known limitation — it needs a real, seekable file to build its matching window against, and cannot read from standard input or write to standard output the way gzip and bzip2 both can.

What compression level does rzip use if I don't pick one?
Level 6 out of a possible 0 (fastest) to 9 (strongest), the same default balance point bzip2 itself uses, though bzip2's own default happens to be its maximum level, 9, rather than a middle setting.