gzip/gunzip for 2.11BSD

A minimal gzip compressor and decompressor implementation for the PDP-11 running 2.11BSD Unix.

Overview

This is a bare-bones implementation of the gzip/DEFLATE compression and decompression algorithms, specifically designed to run on vintage PDP-11 hardware with limited memory. It provides both compression (gzip) and decompression (gunzip) utilities that are compatible with standard gzip files.

Features

gunzip (Decompressor)

• Complete DEFLATE Implementation: Supports all three block types:
  • Uncompressed blocks (type 0)
  • Fixed Huffman coding (type 1)
  • Dynamic Huffman coding (type 2)
• Full gzip Header Parsing: Reads and displays all header information including modification time, flags, original filename, comments, and header CRC
• CRC32 Verification: Validates data integrity by verifying CRC32 checksums
• 32KB Sliding Window: Full DEFLATE-compliant window size
• Progress Indication: Shows decompression progress with percentage and bytes processed
• Stdin/Stdout Support: Read from stdin and/or write to stdout with -c, --stdout, or --to-stdout options
• Standard Compatibility: Decompresses files created by any standard gzip implementation

gzip (Compressor)

• LZ77 String Matching: Hash-based matching with 4KB sliding window
• Memory Efficient: ~12KB total memory usage (4KB window + hash tables)
• Standard gzip Format: Creates files compatible with all gunzip implementations
• CRC32 Checksums: Generates proper CRC32 checksums for data integrity
• Progress Indication: Shows compression progress with percentage and bytes processed
• Stdin/Stdout Support: Read from stdin and/or write to stdout with -c, --stdout, or --to-stdout options
• Pipe-Friendly: Integrates seamlessly in Unix pipelines

Building

make

This will compile both utilities:

• gunzip.c → gunzip executable (decompressor)
• gzip.c → gzip executable (compressor)

Usage

Compressing Files

Compress to a .gz file:

./gzip filename

This creates filename.gz containing the compressed data. The compressor will:

1. Create a standard gzip header with the original filename
2. Compress using LZ77 matching and fixed Huffman encoding
3. Generate CRC32 checksum and write gzip trailer

Example:

$ ./gzip LICENSE
Compressing LICENSE to LICENSE.gz...
Compressed 1083 bytes to 616 bytes

Compress to stdout:

./gzip -c filename > output.gz
./gzip --stdout filename > output.gz
./gzip --to-stdout filename > output.gz

The -c, --stdout, and --to-stdout options write compressed data to stdout instead of creating a .gz file.

Example:

$ ./gzip -c LICENSE > compressed.gz

Compress from stdin:

cat filename | ./gzip > output.gz
./gzip - > output.gz
./gzip < filename > output.gz

When no input filename is specified (or - is used), gzip reads from stdin and writes to stdout. No filename is stored in the gzip header when compressing from stdin.

Example:

$ cat LICENSE | ./gzip > LICENSE.gz
$ echo "Hello, world!" | ./gzip | ./gunzip
Hello, world!

Decompressing Files

Decompress to a file:

./gunzip filename.gz

The decompressor will:

1. Parse and display the gzip header information
2. Decompress the file
3. Write the output to filename (removing the .gz extension)

If the input file doesn't end in .gz, the output will be written to filename.out.

Example:

$ ./gunzip LICENSE.gz
GZIP Header Information:
  Magic:         0x1f 0x8b (valid)
  Method:        8 (deflate)
  Flags:         0x08
    - Original filename present
  Mod time:      1767952414
  Extra flags:   0x04
  OS:            3 (Unix)
  Filename:      LICENSE

Header parsed successfully!
Compressed data starts at byte offset: 21

Decompressing to: LICENSE
Block: final, type=1
Decompression successful! Output: 1083 bytes (CRC OK)

Decompress to stdout:

./gunzip -c filename.gz
./gunzip --stdout filename.gz
./gunzip --to-stdout filename.gz

The -c, --stdout, and --to-stdout options write decompressed data to stdout. When using these options, verbose header information and progress messages are suppressed (only errors go to stderr).

Example:

$ ./gunzip -c LICENSE.gz > LICENSE
$ ./gunzip --stdout file1.gz file2.gz > combined.txt  # Note: only processes first file

Decompress from stdin:

cat filename.gz | ./gunzip
./gunzip - < filename.gz
./gunzip < filename.gz

When no input filename is specified (or - is used), gunzip reads from stdin and writes to stdout.

Example:

$ cat LICENSE.gz | ./gunzip > LICENSE
$ curl https://example.com/file.gz | ./gunzip > file

Pipe Chains

Both utilities support standard Unix pipe workflows:

# Compress and decompress in a pipeline
$ cat file.txt | ./gzip | ./gunzip > output.txt

# Remote backup
$ tar cf - /path/to/backup | ./gzip > backup.tar.gz

# View compressed log without extraction
$ ./gunzip -c logfile.gz | grep ERROR

# Compress multiple files into one archive
$ cat file1 file2 file3 | ./gzip > combined.gz

Implementation Details

Compression Strategy

• Uses fixed Huffman coding (DEFLATE type 1 blocks) for simplicity and compatibility
• 4KB sliding window with hash-based string matching (reduced from 32KB to fit memory constraints)
• Hash chain depth limited to 128 entries for performance
• Minimum match length: 3 bytes
• Maximum match length: 258 bytes

Memory Usage

• gzip: ~12KB total
  • 8KB sliding window (4KB × 2 for circular buffer)
  • 2KB hash table (2048 entries × 2 bytes)
  • 2KB previous links
• gunzip: ~32KB
  • 32KB sliding window (full DEFLATE specification)

Limitations

• Compression is very slow! It takes my (emulated) PDP-11 22 minutes to compress the King James Version of the Bible (4.6MB to 2.3MB) compared to the 6 minutes it takes to decompress the file.
• Single file processing: Processes one file at a time
• Fixed Huffman only for compression: Does not generate dynamic Huffman trees (but achieves reasonable compression ratios)
• Reduced window size for compression: 4KB window vs. standard 32KB (may reduce compression ratio on large files with distant matches)

License

This software is released under the MIT license. See the LICENSE file for license information. Use at your own risk.