Guide

TAR/GZ Archive Format: The Unix Compression Standard Explained

PC By Pablo Cirre • Updated May 7, 2026

TAR/GZ Archive Format: The Unix Compression Standard Explained

TAR (Tape ARchive) is the foundational archive format of Unix and Linux systems, with origins dating to 1979. Unlike ZIP or RAR which combine archiving and compression, TAR separates these concerns: TAR handles file bundling and metadata preservation, while external compression programs (gzip, bzip2, xz, zstd) handle compression. This Unix philosophy approach has made the TAR family of formats — .tar, .tar.gz (.tgz), .tar.bz2, .tar.xz, .tar.zst — the standard for software distribution, system backups, and Linux package management.

TAR File Structure

A TAR file is a sequential format: file headers and data blocks are written one after another with no central index. Each archived file consists of:

[512-byte POSIX header block]
  Filename:      100 bytes (null-terminated, or GNU extended for >100 chars)
  File mode:     8 bytes (octal: 0000755 = rwxr-xr-x)
  UID:           8 bytes (numeric user ID)
  GID:           8 bytes (numeric group ID)
  File size:     12 bytes (octal)
  Modification time: 12 bytes (Unix timestamp, octal)
  Checksum:      8 bytes
  Type flag:     1 byte (0=regular, 2=symlink, 5=directory, 3=char dev, 6=FIFO)
  Link name:     100 bytes (for symlinks)
  [POSIX extension: UStar indicator "ustar", version, uname, gname, device numbers]

[File data, padded to 512-byte boundary]
[Next header block...]
[Two 512-byte zero blocks = end-of-archive marker]

This sequential structure has important implications:

No random access: to extract file #1000, you must read through all 999 preceding headers
Streaming-friendly: TAR can be piped directly to compression and network tools without temporary files
Append only: adding files to an existing TAR requires reading the entire file to find the end marker

TAR Header Formats

Three main header formats exist:

Format	Max filename	Max file size	Key extensions
Original POSIX (v7)	99 chars	8 GB	Basic Unix attributes
UStar (POSIX 1003.1-1988)	255 chars (with prefix)	8 GB	uname, gname, device numbers
GNU tar	Long filenames (LongLink)	Unlimited	Sparse files, volume labels
PAX (POSIX 1003.1-2001)	Unlimited	Unlimited	Extended headers, any attribute

GNU tar and PAX are the formats produced by modern tar implementations on Linux. PAX (--format=posix) is the most portable for archiving very long filenames or very large files.

Compression Algorithms

TAR itself stores no compression — it is a pure archive format. Compression is applied by piping TAR output through a separate compressor:

Extension	Compressor	Flag	Ratio	Speed	Best For
.tar.gz (.tgz)	gzip	-z	Medium	Fast	Universal; best compatibility
.tar.bz2 (.tbz2)	bzip2	-j	Good	Slow	Text/source code archives
.tar.xz (.txz)	xz (LZMA2)	-J	Best	Very Slow	Distribution packages
.tar.zst	zstd	--zstd	Good	Very Fast	Modern Linux distros
.tar.lz4	lz4	(external)	Low	Fastest	Streaming, real-time
.tar.lzo	lzop	(external)	Low	Fast	Embedded systems

# Create archives (all equivalent, different compression)
tar -czf archive.tar.gz directory/        # gzip
tar -cjf archive.tar.bz2 directory/       # bzip2
tar -cJf archive.tar.xz directory/        # xz
tar -c --zstd -f archive.tar.zst directory/  # zstd

# Extract any tar archive (modern tar auto-detects compression)
tar -xf archive.tar.gz
tar -xf archive.tar.bz2
tar -xf archive.tar.xz

# List contents without extracting
tar -tf archive.tar.gz

# Extract single file
tar -xf archive.tar.gz path/to/specific/file.txt

# Create with verbose output
tar -cvzf archive.tar.gz directory/

# Extract to specific directory
tar -xzf archive.tar.gz -C /destination/path/

Preserving Metadata

TAR's key advantage over ZIP for Unix/Linux use is metadata preservation:

# Preserve permissions, ownership, timestamps
tar -cpzf backup.tar.gz /etc/

# Restore with original ownership (requires root)
tar -xpzf backup.tar.gz -C /

# Verify archive integrity
tar -tzf archive.tar.gz > /dev/null && echo "OK"

Preserved metadata includes:

File permissions (read/write/execute for owner/group/other)
Owner and group (by numeric UID/GID in standard TAR, by name in PAX)
Modification timestamps (with nanosecond precision in PAX)
Symbolic links (as link entries, not copies of targets)
Hard links
Device files (character and block devices)
Named pipes (FIFOs)
Extended attributes (SELinux contexts, ACLs in PAX extended headers)

Incremental and Differential Backups

GNU tar supports incremental backups through snapshot files:

# Level 0 (full) backup
tar --listed-incremental=/backup/snapshot.snar \
    -czf /backup/full_$(date +%Y%m%d).tar.gz /home/

# Level 1 (incremental — only files changed since last backup)
tar --listed-incremental=/backup/snapshot.snar \
    -czf /backup/incr_$(date +%Y%m%d).tar.gz /home/

# Restore from incremental chain (full, then each incremental)
tar --listed-incremental=/dev/null -xzf full_20240101.tar.gz
tar --listed-incremental=/dev/null -xzf incr_20240102.tar.gz

Streaming TAR Over SSH and Pipes

TAR's sequential format makes it ideal for streaming:

# Archive and transfer over SSH in one command (no temp file on source)
tar -czf - /data/ | ssh user@remote "cat > /backup/data.tar.gz"

# Transfer directory tree via SSH without creating archive file
tar -cz /source/ | ssh user@remote "tar -xz -C /destination/"

# Archive from remote, extract locally
ssh user@remote "tar -cz /remote/data/" | tar -xz -C /local/data/

# Create archive and immediately verify integrity
tar -czf archive.tar.gz files/ && tar -tzf archive.tar.gz > /dev/null

# Archive entire disk to remote using netcat (faster than SSH for LAN)
# On receiving machine:
nc -l 9000 | tar -xz -C /destination/
# On sending machine:
tar -cz /data/ | nc receiving_host 9000

Compression Comparison for TAR

For a typical mixed-content directory (source code + binaries + data):

Format	Size	Compress Time	Decompress Time
.tar (no compression)	100%	Baseline	Baseline
.tar.gz	~35%	15s	3s
.tar.bz2	~30%	45s	12s
.tar.xz	~25%	120s	5s
.tar.zst	~30%	8s	2s

Recommendation: .tar.zst (Zstandard) offers the best balance for modern systems — compression ratio near bzip2, speed near gzip. .tar.xz is preferred for software distribution where archive size matters more than encoding time.

Linux Package Management and TAR

Most Linux package formats are TAR-based:

Format	Tools	Extension	Compression
Debian/Ubuntu	dpkg, apt	.deb	.tar.xz inside ar
Red Hat/Fedora	rpm, dnf	.rpm	.tar.xz inside CPIO
Arch Linux	pacman	.pkg.tar.zst	.tar.zst
Slackware	pkgtool	.txz	.tar.xz
AppImage	(self-contained)	.AppImage	SquashFS

# Inspect a .deb package contents
dpkg -c package.deb
# or: ar x package.deb && tar -tzf data.tar.xz

# Inspect a .rpm package contents
rpm -qlp package.rpm

TAR vs. ZIP for Different Use Cases

Use Case	Recommendation
Linux/Unix system backups	tar.xz or tar.zst
Software source distribution	tar.xz (standard on GitHub releases)
Windows users sharing	ZIP (native support)
Docker/container layers	tar (OCI image spec uses uncompressed TAR)
Cross-platform sharing	ZIP
Maximum compression	tar.xz
Fast backup streaming	tar.zst

Summary

The TAR family's dominance in Linux/Unix environments stems from its metadata fidelity, streaming capability, and composability with the Unix pipe model. By separating archiving from compression, TAR lets you choose the right compressor for your use case — gzip for universal compatibility, xz for maximum compression, zstd for modern speed. For Windows-primary environments, ZIP's native OS support makes it the practical choice. For any Linux workflow, system backup, or software distribution, TAR with an appropriate compressor remains the correct tool.

Related conversions

Archive format conversions used most often:

Related conversions

Put what you just learned into practice — convert your files now in seconds, free and without registration.

Convert →

Convert →

Convert →

Convert →

Convert →

Convert →

Frequently Asked Questions

.tar is the raw archive with no compression — just files bundled together. .tar.gz (also .tgz) compresses the TAR with gzip: fast, universally supported, ~65% of original size. .tar.bz2 uses bzip2: slower to create but slightly smaller (~30-35%). .tar.xz uses the LZMA2 algorithm (same as 7-Zip): slowest to create, smallest size (~25%), fastest to extract — preferred for software distribution. .tar.zst (Zstandard) is the modern choice: near-bzip2 compression ratio at near-gzip speed.

.tar is the raw archive com no compressão — just arquivos bundled together. .tar.gz (also .tgz) compresses the TAR com gzip: fast, universally suportado, ~65% of original size. .tar.bz2 uses bzip2: slower to create mas slightly smaller (~30-35%). .tar.xz uses the LZMA2 algorithm (same as 7-Zip): slowest to create, smallest size (~25%), fastest to extract — preferred para software distribution. .tar.zst (Zstandard) is the moderno choice: near-bzip2 compressão ratio at near-gzip speed.

.tar is the raw archive mit no Komprimierung — just Dateien bundled together. .tar.gz (also .tgz) compresses the TAR mit gzip: fast, universellly unterstützt, ~65% von original size. .tar.bz2 uses bzip2: slower to create aber slightly smaller (~30-35%). .tar.xz uses the LZMA2 algorithm (same as 7-Zip): slowest to create, smallest size (~25%), fastest to extract — preferred für Software distribution. .tar.zst (Zstandard) is the modern choice: near-bzip2 Komprimierung ratio at near-gzip speed.

.tar is the raw archive con no compresión — just archivos bundled together. .tar.gz (also .tgz) compresses the TAR con gzip: fast, universally soportado, ~65% de original size. .tar.bz2 uses bzip2: slower to create pero slightly smaller (~30-35%). .tar.xz uses the LZMA2 algorithm (same as 7-Zip): slowest to create, smallest size (~25%), fastest to extract — preferred para software distribution. .tar.zst (Zstandard) is the moderno choice: near-bzip2 compresión ratio at near-gzip speed.

7z (LZMA2) typically beats both ZIP and RAR by 10–30% for text-heavy content, but is slower to create. RAR compresses slightly better than ZIP and supports recovery records (good for damaged archives) but the format is proprietary. ZIP is the universal lingua franca — every OS opens it natively. Use 7z for long-term storage, ZIP for sharing.

Use: tar -xzf archive.tar.gz path/to/specific/file.txt — providing the exact path as it appears inside the archive. First list the archive contents to find the exact path: tar -tzf archive.tar.gz | grep filename. Note the path includes any directory prefix — if the archive was created with tar -cz mydir/, files will be listed as mydir/filename, so extract with: tar -xzf archive.tar.gz mydir/filename.txt.

Use: tar -xzf archive.tar.gz path/to/specific/file.txt — providing the exact path as it appears inside the archive. First list the archive contents to find the exact path: tar -tzf archive.tar.gz | grep filename. Note the path inclui any directory prefix — if the archive was created com tar -cz mydir/, arquivos will be listed as mydir/filename, so extract with: tar -xzf archive.tar.gz mydir/filename.txt.

Use: tar -xzf archive.tar.gz path/to/specific/file.txt — providing the exact path as it appears inside the archive. First list the archive contents to find the exact path: tar -tzf archive.tar.gz | grep filename. Note the path beinhaltet any directory prefix — if the archive was created mit tar -cz mydir/, Dateien will be listed as mydir/filename, so extract with: tar -xzf archive.tar.gz mydir/filename.txt.

Use: tar -xzf archive.tar.gz path/to/specific/file.txt — providing the exact path as it appears inside the archive. First list the archive contents to find the exact path: tar -tzf archive.tar.gz | grep filename. Note the path incluye any directory prefix — if the archive was created con tar -cz mydir/, archivos will be listed as mydir/filename, so extract with: tar -xzf archive.tar.gz mydir/filename.txt.

Yes, when the file contains anything sensitive. Standard ZIP "encryption" (ZipCrypto) is broken and trivially crackable. Use AES-256 (ZIP 2.0+, supported by 7-Zip, WinRAR, modern Windows). Verify the encryption type when creating: many tools default to ZipCrypto for compatibility. Better still, use 7z which uses AES-256 by default.

TAR was designed specifically for Unix systems where file permissions, ownership, and special file types (symlinks, device files, FIFOs) are fundamental. Its 512-byte header explicitly stores Unix mode bits, UID, GID, and file type flags. ZIP was designed for cross-platform use (DOS origin) with no concept of Unix permissions. Most ZIP implementations add Unix permissions as extra fields, but these are not standardized and may be ignored by Windows tools. For Unix system backups, TAR with -p (preserve permissions) is the correct choice.

TAR was designed specifically para Unix systems where arquivo permissions, ownership, e special tipo de arquivos (symlinks, device files, FIFOs) are fundamental. Its 512-byte header explicitly stores Unix mode bits, UID, GID, e tipo de arquivo flags. ZIP was designed para cross-platform usar (DOS origin) com no concept of Unix permissions. Most ZIP implementations add Unix permissions as extra fields, mas these are not padrãoized e may be ignored by Windows ferramentas. para Unix system backups, TAR com -p (preserve permissions) is the correct choice.

TAR was designed specifically für Unix systems where Datei permissions, ownership, und special Dateityps (symlinks, device files, FIFOs) are fundamental. Its 512-byte header explicitly stores Unix mode bits, UID, GID, und Dateityp flags. ZIP was designed für cross-platform verwenden (DOS origin) mit no concept von Unix permissions. Most ZIP implementations add Unix permissions as extra fields, aber these are not Standardized und may be ignored by Windows Werkzeuge. für Unix system backups, TAR mit -p (preserve permissions) is the correct choice.

TAR was designed specifically para Unix systems where archivo permissions, ownership, y special tipo de archivos (symlinks, device files, FIFOs) are fundamental. Its 512-byte header explicitly stores Unix mode bits, UID, GID, y tipo de archivo flags. ZIP was designed para cross-platform usar (DOS origin) con no concept de Unix permissions. Most ZIP implementations add Unix permissions as extra fields, pero these are not estándarized y may be ignored by Windows herramientas. para Unix system backups, TAR con -p (preserve permissions) is the correct choice.

Original ZIP: 4 GB total + 65,535 files. ZIP64 (transparent extension): up to 16 EB and unlimited files. Modern tools (Windows Explorer, macOS Finder, 7-Zip, Linux unzip) all support ZIP64 automatically when needed. The only catch: very old tools (pre-2007) cannot open ZIP64 archives.

Modern GNU tar (1.31+) includes native Zstandard support via --zstd: tar --zstd -cf archive.tar.zst directory/. On systems with older tar, pipe manually: tar -cf - directory/ | zstd -o archive.tar.zst. To extract: tar --zstd -xf archive.tar.zst or tar -xf archive.tar.zst (auto-detection). Set compression level with ZSTD_CLEVEL env variable or: tar -cf - directory/ | zstd -19 -o archive.tar.zst (level 19 = maximum compression).

Modern GNU tar (1.31+) inclui native Zstandard support via --zstd: tar --zstd -cf archive.tar.zst directory/. On systems com older tar, pipe manually: tar -cf - directory/ | zstd -o archive.tar.zst. To extract: tar --zstd -xf archive.tar.zst ou tar -xf archive.tar.zst (auto-detection). Set compressão level com ZSTD_CLEVEL env variable or: tar -cf - directory/ | zstd -19 -o archive.tar.zst (level 19 = máximo compressão).

Modern GNU tar (1.31+) beinhaltet native Zstandard support via --zstd: tar --zstd -cf archive.tar.zst directory/. On systems mit older tar, pipe manually: tar -cf - directory/ | zstd -o archive.tar.zst. To extract: tar --zstd -xf archive.tar.zst oder tar -xf archive.tar.zst (auto-detection). Set Komprimierung level mit ZSTD_CLEVEL env variable or: tar -cf - directory/ | zstd -19 -o archive.tar.zst (level 19 = maximal Komprimierung).

Modern GNU tar (1.31+) incluye native Zstandard support via --zstd: tar --zstd -cf archive.tar.zst directory/. On systems con older tar, pipe manually: tar -cf - directory/ | zstd -o archive.tar.zst. To extract: tar --zstd -xf archive.tar.zst o tar -xf archive.tar.zst (auto-detection). Set compresión level con ZSTD_CLEVEL env variable or: tar -cf - directory/ | zstd -19 -o archive.tar.zst (level 19 = máximo compresión).

Some content does not compress well: already-compressed media (JPG, MP4, MP3), encrypted files, random binary data. Compression algorithms exploit redundancy — these formats already removed it. To shrink them you would need lossy re-encoding, not archiving. Save archives for text, source code, uncompressed media (BMP, WAV, RAW), and document files.