MSZIP Compression

Purpose

This guide explains MSZIP (Microsoft ZIP) compression, the default compression algorithm for CAB files based on DEFLATE.

References

Concepts

What is MSZIP?

MSZIP is Microsoft’s implementation of the DEFLATE algorithm, the same compression used in ZIP files. It provides a good balance between compression ratio and speed.

Key Features

  • DEFLATE-based - Compatible with ZIP compression

  • 32 KB window - Larger than LZSS (4 KB)

  • Huffman coding - Statistical encoding for better compression

  • Block-based - Compresses data in 32 KB blocks (CAB)

Performance characteristics

Metric Value

Compression ratio

40-60% for text, 30-50% for executables

Compression speed

Medium (5-20 MB/s)

Decompression speed

Fast (20-100 MB/s)

Memory usage

Medium (32 KB)

Window size

32,768 bytes (32 KB)

Usage

CAB Files (Default) 

compressor = Cabriolet::CAB::Compressor.new(compression: :mszip)
compressor.add_file('data.txt')
compressor.write('archive.cab')

Command line:

# MSZIP is default
cabriolet create archive.cab file1.txt file2.txt

# Explicit MSZIP
cabriolet create --compression=mszip archive.cab data.txt

KWAJ Files 

compressor = Cabriolet::KWAJ::Compressor.new

bytes = compressor.compress(
  'data.bin',
  'data.kwj',
  compression: :mszip
)

Algorithm Details

Two-Stage Compression

  1. LZ77 Dictionary Compression

    • Find repeated sequences

    • Encode as (distance, length) pairs

    • 32 KB sliding window

  2. Huffman Encoding

    • Build frequency tables

    • Create optimal prefix codes

    • Encode literals and matches

Block Structure

CAB uses 32 KB blocks, each prefixed with a "CK" signature:

CFDATA Block 1:
  [CK signature] [DEFLATE blocks...] → 32 KB decompressed

CFDATA Block 2:
  [CK signature] [DEFLATE blocks...] → 32 KB decompressed

CFDATA Block 3:
  [CK signature] [DEFLATE blocks...] → Remaining bytes

Multi-File Extraction

When multiple files share a CFDATA block, Cabriolet maintains window state:

CFDATA Block (32 KB decompressed):
  ├── file1.txt (10 KB) → Written to disk
  ├── file2.txt (15 KB) → Written to disk
  └── file3.txt (7 KB)  → Written to disk

Window buffer preserves unconsumed data between extract calls.

This is handled automatically by the decompressor’s @window_offset tracking.

Compression Effectiveness

Excellent For

  • Text files - 50-60% compression

  • Source code - 55-65% compression

  • HTML/XML - 55-65% compression

  • Log files - 60-70% compression

Good For

  • Executables - 40-50% compression

  • DLLs - 40-50% compression

  • Office documents - 30-40% compression

Poor For

  • Already compressed - 0-5% (JPG, PNG, MP3, ZIP)

  • Random data - 0-10%

  • Encrypted files - 0-5%

Comparison

Algorithm Text Ratio Speed Memory

LZSS

40-50%

Fast

4 KB

MSZIP

50-60%

Medium

32 KB

LZX

60-70%

Slow

32KB-2MB

Quantum

50-65%

Medium

Variable

Best practices

  1. Default choice - Use for general-purpose CAB files

  2. Mixed content - Good for varied file types

  3. Distribution - Balance of size and speed

  4. Compatibility - Widely supported

  5. Skip media - Don’t compress JPG, PNG, MP4

Examples

Create Installation Package 

require 'cabriolet'

compressor = Cabriolet::CAB::Compressor.new(compression: :mszip)

# Add application files
compressor.add_file('setup.exe')
compressor.add_file('readme.txt')
compressor.add_file('license.txt')

compressor.write('installer.cab')
puts "Created installer with MSZIP compression"

Mixed Compression 

require 'cabriolet'

# Folder 1: MSZIP for most files
mszip_compressor = Cabriolet::CAB::Compressor.new(compression: :mszip)
mszip_compressor.add_file('program.exe')
mszip_compressor.add_file('config.xml')

# Folder 2: None for pre-compressed
none_compressor = Cabriolet::CAB::Compressor.new(compression: :none)
none_compressor.add_file('logo.jpg')
none_compressor.add_file('music.mp3')

Compression Statistics 

require 'cabriolet'

compressor = Cabriolet::CAB::Compressor.new(compression: :mszip)

original_size = 0
Dir.glob('*.txt').each do |file|
  compressor.add_file(file)
  original_size += File.size(file)
end

compressed_size = compressor.write('archive.cab')

ratio = (1 - compressed_size.to_f / original_size) * 100
puts "Compression ratio: #{ratio.round(1)}%"
puts "#{original_size} → #{compressed_size} bytes"

See also

Bibliography