SZDD Format Guide

Purpose

This guide provides comprehensive documentation for working with Microsoft SZDD (Single-file LZSS) compressed files using Cabriolet. SZDD is Microsoft’s simple compression format used primarily in MS-DOS and early Windows for compressing individual files.

References

• Getting Started Guide

• SZDD CLI Commands

• LZSS Compression Guide

• Ruby API Reference

Concepts

What is an SZDD File?

SZDD files are Microsoft’s single-file compression format used extensively in:

• MS-DOS file compression (COMPRESS.EXE)

• MS-DOS file expansion (EXPAND.EXE)

• Windows 9x installation files

• System file distribution

• Compressed executables and DLLs

The format is characterized by its simple structure and use of LZSS compression.

SZDD File Structure

An SZDD file has a simple structure:

┌─────────────────────────┐
│   SZDD Header           │  Signature and metadata (14 bytes)
│   - Signature: "SZDD"   │
│   - Format version      │
│   - Missing character   │
│   - Uncompressed size   │
├─────────────────────────┤
│   Compressed Data       │  LZSS compressed content
│                         │
└─────────────────────────┘

SZDD Header: Contains format signature, version, and optional metadata.

Missing Character: A special feature for filename reconstruction (e.g., file.ex_ → file.exe).

Compressed Data: LZSS-compressed file content.

Compression Support

SZDD files use LZSS compression exclusively:

• LZSS MODE_EXPAND - Standard LZSS variant used by EXPAND.EXE

• Fixed 4096-byte sliding window

• Simple and fast decompression

For detailed algorithm information, see link:.

Format Variants

SZDD has two format variants:

• Normal format - Standard SZDD with signature "SZDD"

• QBasic format - Variant used by QBasic with signature "SZ20"

Both use the same compression but differ in header structure.

Filename Convention

SZDD files traditionally use underscore replacement for the last character:

• file.txt → file.tx_

• program.exe → program.ex_

• library.dll → library.dl_

The missing character is stored in the header for reconstruction.

Basic Operations

Expanding SZDD Files

Decompress SZDD files (like MS-DOS EXPAND.EXE):

Command-line

# Auto-detect output filename
cabriolet extract file.tx_

# Specify output filename
cabriolet extract file.tx_ file.txt

# For explicit format specification:
cabriolet extract --format szdd file.tx_ file.txt

The expand command is a legacy alias for extract and is still supported for backward compatibility with MS-DOS EXPAND.EXE.

Ruby API

require 'cabriolet'

decompressor = Cabriolet::SZDD::Decompressor.new
header = decompressor.open('file.tx_')

# Auto-detect output name from missing character
output = decompressor.auto_output_filename('file.tx_', header)

# Expand the file
bytes = decompressor.extract(header, output)
decompressor.close(header)

puts "Expanded to #{output} (#{bytes} bytes)"

Example Output

Expanded file.tx_ to file.txt (5,234 bytes)

Getting SZDD Information

Display SZDD file metadata:

Command-line

cabriolet info file.tx_

# For explicit format specification:
cabriolet info --format szdd file.tx_

Ruby API

require 'cabriolet'

decompressor = Cabriolet::SZDD::Decompressor.new
header = decompressor.open('file.tx_')

puts "Format: #{header.format}"  # :normal or :qbasic
puts "Uncompressed size: #{header.length} bytes"
puts "Missing character: '#{header.missing_char}'"
puts "Suggested filename: #{header.suggested_filename('file.tx_')}"

decompressor.close(header)

Example Output

SZDD File Information
==================================================
Filename: file.tx_
Format: NORMAL
Uncompressed size: 5,234 bytes
Missing character: 't'
Suggested filename: file.txt

Compressing Files

Create SZDD compressed files (like MS-DOS COMPRESS.EXE):

Command-line

# Auto-generate output name: file.txt → file.tx_
cabriolet create file.tx_ file.txt

# Specify output filename
cabriolet create file.tx_ file.txt

# Specify missing character explicitly
cabriolet create --missing-char=t file.tx_ file.txt

# Use QBasic format
cabriolet create --format=qbasic file.ba_ file.bas

# For explicit format specification:
cabriolet create --format szdd file.tx_ file.txt

The compress command is a legacy alias for create and is still supported for backward compatibility with MS-DOS COMPRESS.EXE.

Ruby API

require 'cabriolet'

compressor = Cabriolet::SZDD::Compressor.new

# Compress with auto-generated missing character
bytes = compressor.compress('file.txt', 'file.tx_')

puts "Compressed to file.tx_ (#{bytes} bytes)"

Advanced Features

Missing Character Reconstruction

The missing character feature enables filename reconstruction:

require 'cabriolet'

decompressor = Cabriolet::SZDD::Decompressor.new
header = decompressor.open('program.ex_')

# Get missing character
missing = header.missing_char  # => 'e'

# Reconstruct original filename
original = header.suggested_filename('program.ex_')
puts "Original filename: #{original}"  # => "program.exe"

decompressor.close(header)

Format detection

Detect SZDD format variant:

require 'cabriolet'

decompressor = Cabriolet::SZDD::Decompressor.new
header = decompressor.open('file.tx_')

case header.format
when :normal
  puts "Standard SZDD format"
when :qbasic
  puts "QBasic SZDD variant"
end

decompressor.close(header)

Custom Missing Character

Specify custom missing character when compressing:

require 'cabriolet'

compressor = Cabriolet::SZDD::Compressor.new

# Compress with custom missing character
bytes = compressor.compress(
  'myfile.txt',
  'myfile.tx_',
  missing_char: 't'
)

puts "Compressed with missing char 't'"

QBasic Format Support

Work with QBasic-compressed files:

require 'cabriolet'

# Compress in QBasic format
compressor = Cabriolet::SZDD::Compressor.new
bytes = compressor.compress(
  'program.bas',
  'program.ba_',
  format: :qbasic
)

# Decompress QBasic format
decompressor = Cabriolet::SZDD::Decompressor.new
header = decompressor.open('program.ba_')

if header.format == :qbasic
  puts "QBasic format detected"
end

decompressor.extract(header, 'program.bas')
decompressor.close(header)

Performance Optimization

Compression Trade-offs

SZDD uses LZSS which offers:

Characteristic	Details
Compression speed	Fast (simple algorithm)
Decompression speed	Very fast
Compression ratio	Moderate (30-50% typical)
Memory usage	Low (4KB window)

Best File Types

SZDD works best for:

• Text files - Good compression (40-60%)

• Executables - Moderate compression (30-40%)

• HTML/XML - Good compression (50-70%)

Not recommended for:

• Already compressed - JPG, PNG, ZIP (no benefit)

• Very small files - <1KB (overhead too high)

Common Use Cases

MS-DOS File Distribution

Working with MS-DOS compressed files:

require 'cabriolet'

# Expand MS-DOS system file
decompressor = Cabriolet::SZDD::Decompressor.new
header = decompressor.open('command.co_')

output = decompressor.auto_output_filename('command.co_', header)
decompressor.extract(header, output)
decompressor.close(header)

puts "Expanded MS-DOS file: #{output}"

Windows 9x Installation Files

Extract Windows installation files:

require 'cabriolet'
require 'fileutils'

decompressor = Cabriolet::SZDD::Decompressor.new

# Process all ._ files in directory
Dir.glob('install/*.ex_').each do |compressed|
  header = decompressor.open(compressed)
  output = decompressor.auto_output_filename(compressed, header)

  decompressor.extract(header, output)
  decompressor.close(header)

  puts "Extracted: #{File.basename(output)}"
end

Batch Compression

Compress multiple files:

require 'cabriolet'

compressor = Cabriolet::SZDD::Compressor.new

Dir.glob('source/*.txt').each do |file|
  # Generate compressed filename: file.txt → file.tx_
  output = file.sub(/\.([^.])$/, '._')

  bytes = compressor.compress(file, output)
  puts "Compressed #{file} → #{output} (#{bytes} bytes)"
end

Filename Reconstruction

Reconstruct original filenames:

require 'cabriolet'

decompressor = Cabriolet::SZDD::Decompressor.new

Dir.glob('compressed/*._').each do |compressed|
  header = decompressor.open(compressed)

  # Use missing character to reconstruct name
  suggested = header.suggested_filename(compressed)

  puts "#{File.basename(compressed)} → #{File.basename(suggested)}"
  puts "  Missing char: '#{header.missing_char}'"

  decompressor.close(header)
end

Troubleshooting

Common Errors

"Invalid SZDD signature"

The file is not a valid SZDD file:

# Check file signature
hexdump -C file.tx_ | head -1
# Should show: "SZDD" (53 5A 44 44) or "SZ20" for QBasic

"Missing character not found"

The header doesn’t contain a missing character. Specify output filename explicitly:

cabriolet extract file.tx_ file.txt

"Decompression failed"

The compressed data is corrupted. Try salvage extraction:

# Extract partial data (planned feature)
cabriolet extract --salvage file.tx_ file.txt

"Unknown format variant"

Unsupported SZDD variant. Check format with:

cabriolet info file.tx_

Validation

Verify SZDD file integrity:

require 'cabriolet'

begin
  decompressor = Cabriolet::SZDD::Decompressor.new
  header = decompressor.open('file.tx_')

  puts "SZDD file is valid"
  puts "Format: #{header.format}"
  puts "Size: #{header.length} bytes"

  decompressor.close(header)
rescue Cabriolet::FormatError => e
  puts "Invalid SZDD file: #{e.message}"
rescue Cabriolet::CorruptionError => e
  puts "Corrupted SZDD: #{e.message}"
end

Best practices

1. Use missing character: Store the missing character for proper filename reconstruction

2. Match format to use case:

   • Normal format for general files

   • QBasic format for QBasic programs

3. Validate before extraction: Check file integrity before expanding

4. Preserve metadata: Keep track of original filenames and sizes

5. Batch processing: Process multiple files efficiently with scripts

6. Handle errors gracefully: Account for corrupted or invalid files

Format Specifications

File Signature

SZDD files start with one of two signatures:

Normal Format

Offset  Bytes  Description
0x0000  4      Signature: "SZDD" (0x53 0x5A 0x44 0x44)
0x0004  4      Reserved (0x88 0xF0 0x27 0x33)
0x0008  1      Compression mode: 'A' (0x41)
0x0009  1      Missing character (ASCII)
0x000A  4      Uncompressed length (little-endian)

QBasic Format

Offset  Bytes  Description
0x0000  4      Signature: "SZ20" (0x53 0x5A 0x32 0x30)
0x0004  4      Reserved
0x0008  1      Compression mode: 'A' (0x41)
0x0009  1      Missing character
0x000A  4      Uncompressed length

Compression Algorithm

• Algorithm: LZSS MODE_EXPAND

• Window size: 4096 bytes

• Match length: 3-18 bytes

• Format: See LZSS Compression Guide

For complete format specifications, see Format Specifications.

Next steps

• Learn about LZSS compression

• Review extracting files

• Study Ruby API

See also

• KWAJ Format

• CAB Format

• All Format Guides

Bibliography

• Cabriolet on GitHub

• MS-DOS EXPAND.EXE Documentation

• Wikipedia: SZDD

• cabextract Documentation