Multi-Volume Archives

This document specifies multi-volume (split) archive handling in 7z format.

Overview

Multi-volume archives split a single logical archive across multiple physical files (volumes). This enables:

• Distribution on size-limited media
• Easier transfer of large archives
• Incremental download/upload

Naming Convention

Volumes use zero-padded numeric extensions:

archive.7z.001    # First volume
archive.7z.002    # Second volume
archive.7z.003    # Third volume
...
archive.7z.NNN    # Nth volume

Extension format: .NNN (always 3 digits, zero-padded)

Maximum volumes: 999 (extensions 001-999)

Beyond 999: Extensions .1000 and higher are undefined by this specification. Implementations encountering such extensions SHOULD treat them as continuation volumes using implementation-defined behavior, or MAY reject the archive. Writers MUST NOT create archives with more than 999 volumes.

Volume Structure

First Volume

Contains the signature header:

┌─────────────────────────┐
│    Signature Header     │  32 bytes
├─────────────────────────┤
│    Pack Data (start)    │  Up to volume boundary
└─────────────────────────┘

Middle Volumes

Contain only pack data:

┌─────────────────────────┐
│  Pack Data (continued)  │  Full volume
└─────────────────────────┘

Last Volume

Contains pack data end and header:

┌─────────────────────────┐
│   Pack Data (end)       │  Remaining compressed data
├─────────────────────────┤
│   Next Header           │  Archive metadata
└─────────────────────────┘

Virtual Stream

Volumes concatenate to form a virtual stream:

Volume 1:  [Signature][Data 0x00-0x0FFF]
Volume 2:  [Data 0x1000-0x1FFF]
Volume 3:  [Data 0x2000-0x27FF][Header]

Virtual:   [Signature][Data 0x00-0x27FF][Header]

Header Location

The header is always in the last volume:

1. NextHeaderOffset in signature header points past all pack data
2. Seeking to header position requires seeking into appropriate volume
3. Header must fit entirely within the last volume

Volume Size Calculation

When creating multi-volume archives:

function write_volumes(data, header, volume_size):
    total_data = len(data) + len(header)

    # First volume has 32-byte signature
    first_volume_data = volume_size - 32
    remaining = total_data - first_volume_data

    # Calculate needed volumes
    middle_volumes = remaining / volume_size
    last_volume_size = remaining % volume_size

    # Ensure header fits in last volume
    if last_volume_size < len(header):
        # Adjust distribution

Reading Multi-Volume Archives

Discovery

1. Open first volume (*.001)
2. Read signature header
3. Discover subsequent volumes by incrementing extension
4. Verify all volumes exist and have expected sizes

Virtual Seek

function seek(position):
    if position < 32:
        # Within signature header (volume 1)
        return volume[0].seek(position)

    data_position = position - 32
    volume_index = 0
    remaining = data_position

    # Skip past first volume's data portion
    first_data_size = volume_sizes[0] - 32
    if remaining < first_data_size:
        return volume[0].seek(32 + remaining)
    remaining -= first_data_size
    volume_index = 1

    # Find correct middle/last volume
    while volume_index < len(volumes):
        vol_size = volume_sizes[volume_index]
        if remaining < vol_size:
            return volume[volume_index].seek(remaining)
        remaining -= vol_size
        volume_index += 1

    error("Seek past end of archive")

Virtual Read

function read(position, size):
    result = []
    while len(result) < size:
        volume, offset = translate_position(position + len(result))
        available = volume.size - offset
        chunk = min(size - len(result), available)
        result.extend(volume.read(offset, chunk))
    return result

Writing Multi-Volume Archives

Process

1. Determine volume size
2. Write signature header to volume 1
3. Compress and write pack data, splitting across volumes
4. Write header to last volume
5. Finalize signature header offsets

Volume Boundary Handling

Pack streams may span volume boundaries:

Pack Stream: [AAAA|BBBB|CCCC]
                  ^    ^
             Vol boundary

Volume 1: [Sig][AAAA]
Volume 2: [BBBB]
Volume 3: [CCCC][Header]

This is transparent to the compression layer.

Incomplete Volume Sets

Detection

Check for missing volumes:

• Verify sequential numbering (001, 002, 003...)
• Verify each volume size (except last may differ)
• Verify total size matches signature header expectations

Error Handling

Condition	Action
Missing volume	Abort with error indicating which volume
Truncated volume	Abort with corruption error
Extra files	Ignore (may be backup copies)
Wrong order	Volume numbers ensure correct ordering

Atomic Operations

Updating multi-volume archives:

1. Cannot modify in-place (would require shifting all volumes)
2. Create new volume set for updates
3. Delete old volumes only after new set is complete

Size Recommendations

Media	Volume Size
CD-R	700 MiB
DVD	4.7 GiB
FAT32	< 4 GiB
Cloud upload	100-500 MiB
Email attachment	10-25 MiB

Implementation Notes

Memory Mapping

For large volumes, use memory-mapped I/O or streaming reads rather than loading entire volumes.

Parallel Processing

Volumes can be read in parallel when extracting from non-solid archives (if files are in different volumes).

Network Resilience

For network transfers:

• Verify each volume checksum before processing
• Support resumable downloads per-volume
• Allow partial extraction from available volumes

Example

Archive split into 3 volumes of 10 MiB each:

archive.7z.001:
  Offset 0x00: Signature Header
  Offset 0x20: Pack data bytes 0-10485727

archive.7z.002:
  Offset 0x00: Pack data bytes 10485728-20971455

archive.7z.003:
  Offset 0x00: Pack data bytes 20971456-25000000
  Offset after pack: Next Header (remaining space)

Signature header contains:

• NextHeaderOffset = 25000001 - 32 = 24999969
• NextHeaderSize = (size of header)

Compatibility

• 7-Zip: Full support
• p7zip: Full support
• Most third-party tools: Support varies

Some tools only support single-file archives; always test compatibility for target environment.

See Also

• Archive Structure - Single-volume layout
• Signature Header - Header offset interpretation
• Error Conditions - Volume error handling