TZ File — What a Tar Archive Compressed With the Old Unix Compress Format Is

A TZ File Bundles Tar's Structure With Compress's Older Algorithm

A .tz file is a tar archive — files bundled together using the 512-byte header records tar has used since Seventh Edition Unix in 1979 — compressed with the Unix compress utility rather than gzip or bzip2. Compress uses LZW compression, first implemented by Spencer Thomas in 1984 and finalized as version 4.0 in 1985 by Joseph Orost and collaborators, which predates gzip, bzip2, and every archive compression format built after the mid-1990s. The exact same file is also seen with the extensions .tar.Z and .taz, all three representing an identical tar-plus-compress combination with no technical difference between them whatsoever.

Every compress-produced stream, including the one wrapped inside a .tz file, begins with a fixed two-byte magic number, 0x1F 0x9D, immediately followed by a flag byte recording the maximum code width used during that particular compression run. A decompressor checks that magic number first and refuses to proceed if it doesn't match, which is why a .tz file can't simply be renamed from some other format and expected to decompress correctly.

That maximum code width is a real, adjustable setting, not a fixed constant. The POSIX specification for the compress utility documents its -b option as accepting a range of 9 to 16 bits, with the default allowed to be 14, 15, or 16 depending on the implementation. LZW starts encoding at 9-bit codes and grows its dictionary of previously seen byte sequences adaptively, widening the code size step by step until it hits whichever maximum the -b setting allows. A separate -c option sends the compressed output to standard output instead of a new file, which is what lets compress be chained directly after tar in a single pipeline.

Classic compress also behaves differently from most later compressors once its dictionary fills up: the algorithm monitors its own compression ratio as it works, and when that ratio drops enough, it clears the entire dictionary and starts rebuilding it from scratch partway through the file. A correct decompressor has to detect and replay that same reset at the identical point in the stream, or everything decoded after it comes out wrong — a real, documented quirk of LZW's design that later algorithms like DEFLATE and LZMA2 don't share in the same way.

Why the Extension Got Shortened, and Why Compress Itself Was Abandoned

The shortened .tz spelling exists for the same practical reason .tgz and .tbz2 do: early MS-DOS and Windows systems enforced an 8.3 filename rule, and some early Unix variants historically limited filenames to just 14 characters total, both of which made "archive.tar.Z" awkward to work with. Compressing to .tz instead of the full ".tar.Z" fits comfortably within either restriction.

Compress itself has a documented reason it stopped being the default choice for new archives: its LZW algorithm was covered by two U.S. patents, 4,464,650 (issued August 7, 1984, covering the underlying Lempel-Ziv algorithm) and 4,558,302 (issued December 10, 1985, covering the specific Lempel-Ziv-Welch variant), both originally assigned to Sperry Corporation and later enforced by Unisys after Sperry merged into it. In the early 1990s Unisys reversed an earlier informal royalty-free stance and began actively enforcing licensing terms around LZW, most visibly against the GIF image format. Gzip was released on October 31, 1992 by Jean-loup Gailly and Mark Adler specifically as a free, patent-unencumbered replacement for compress, and it quickly displaced compress as the default across nearly every Unix-like system, years before both LZW patents eventually expired.

That patent history is also why GNU tar's own documentation still treats compress as a legacy option. The GNU tar manual, current as of version 1.35, describes compress as supported "mostly for backward compatibility" and recommends against using it, calling it "by far less effective" than tar's other compression programs — a direct statement from the tool's own maintainers about why compress lost its place as the default.

LZW's dictionary-based approach, which starts encoding at 9-bit codes and grows adaptively up to a maximum of 16 bits as its dictionary of seen byte sequences fills, is also structurally different from DEFLATE's LZ77-plus-Huffman design and generally compresses less tightly on the same data, giving gzip a real ratio advantage on top of the patent-free one.

What Encountering a TZ File Today Actually Means

• Gain — compatibility with genuinely old decompression tools: some Unix systems and toolchains from the 1980s and 1990s can decompress .tz but were never updated to understand gzip or newer formats.
• Lose — the compression ratio gzip or newer formats would provide: LZW's simpler dictionary coding generally compresses less tightly than DEFLATE, bzip2, or LZMA2 on the same input.
• Unchanged — every file's name, permissions, and content once extracted: tar's own header structure and compress's lossless algorithm mean nothing about the archived data is altered or approximated.
• Lose — the patent concern that originally mattered: the LZW patent expired in 2003, so while compress is obsolete in practice, using it carries no legal risk today.
• Gain — a strong signal about the file's age or origin: encountering a genuinely new .tz file today almost always points to a legacy pipeline, archival project, or specialized system still built around the older format.

Which Current Tools Can Still Open a TZ File

Modern operating systems generally still support reading .tz files even though very few tools default to creating new ones. On Linux, the ncompress package and certain gzip distributions retain a compress-compatible decompression mode. Apple's own documented Archive Utility feature list on macOS names the lowercase .z extension among the formats it opens natively, though .tz and .taz specifically aren't listed by name in that same feature list, which is a real, documented gap worth checking before assuming Archive Utility will handle a .tz file exactly the way it handles .tgz or .tbz2.

On Windows, 7-Zip and other third-party archive tools generally focus on extraction rather than creation for this specific format, treating .tz largely as a read-only legacy format at this point. 7-Zip's own documented list of supported formats places Z (Unix compress) specifically among the formats it can only extract and never create, confirming that a .tz file can be opened in 7-Zip but never rebuilt from scratch there. Windows 10 version 1803 and later ships a command-line tar.exe based on libarchive, and libarchive's own documentation lists compress among the compression formats it can decode, giving Windows a command-line path to open a .tz file without installing a separate archive utility.

On the creation side, rebuilding a genuinely new .tz file from scratch is comparatively rare in current software. The ncompress package, a free reimplementation of the original compress program, remains the most direct way to produce compress-compatible output on Linux, since many current distributions no longer install the classic compress binary by default. GNU tar's own -Z / --compress option still performs this step directly from a single command as of GNU tar 1.35, even though the manual itself discourages relying on it for new archives.

Why Someone Still Runs Into a TZ File in 2026

A documented, recurring reason is an older scientific, geospatial, or industrial dataset pipeline built decades ago around compress that was never fully modernized, where downstream scripts or specialized hardware firmware still expect exactly this extension and encoding rather than a newer compression format. Replacing the entire pipeline is often riskier or more expensive than continuing to produce files in the format it already expects.

A second real pattern involves archival preservation projects deliberately handling files in their original historical format, including compress output, specifically to document how period-appropriate software processed that data. A third scenario involves cross-platform scripts originally written for older Unix systems that call the compress binary directly by name rather than a general-purpose library — encountering a .tz file from one of these scripts is a sign the underlying automation hasn't yet been modernized to call gzip or another current tool instead.

TZ's LZW Compression Set Beside the Formats That Replaced It

Common Questions About What a TZ File Is

Is a .tz file the same thing as a .tar.Z file?
Yes, completely, and .taz is also the same format under a third spelling. All three describe an identical tar archive compressed with the Unix compress utility's LZW algorithm.

Why don't more tools create new TZ files anymore?
Gzip and newer formats generally compress tighter and were adopted specifically to move away from compress's LZW algorithm, which carried patent licensing concerns through 2003 and has been the default replacement target ever since.

Does my Mac open a TZ file the same way it opens a TGZ file?
Not necessarily. Apple's documented Archive Utility feature list names .z but doesn't specifically list .tz or .taz the way it lists .tgz and .tbz2, so a separate tool may be needed depending on the exact file.

Is there any data loss risk in decompressing an old TZ file?
No. LZW compression, like gzip's DEFLATE, is lossless, so decompressing a TZ file restores the exact original tar bytes with nothing discarded.

Is compress's LZW algorithm still covered by patents today?
No. The two U.S. patents that once covered it, 4,464,650 and 4,558,302, both originally assigned to Sperry Corporation and later enforced by Unisys, have long since expired, so there's no remaining licensing concern with the format itself.

Can 7-Zip create a new TZ file, since it can already open one?
No. 7-Zip's own documented supported-formats list places Z among the formats it can only extract, not create, so it opens an existing .tz file without issue but has no option to build a new one from scratch.