ARJ file format: headers and codes

An ARJ archive is a main header, then a local file header and the compressed data for each file, then an end marker. Both kinds of header have the same frame and share most of their first 34 bytes. This page follows Robert Jung’s technical note TECHNOTE.DOC (the 1994 edition shipped with ARJ 2.4x); fields added by later ARJ versions are marked as such and come from the open-source ARJ and the parsers built on it. Numbers are little-endian.

Header frame

Offset	Size	Name	Description
`0`	2	Header ID	`60 EA` (the number `EA60`).
`2`	2	Basic header size	Bytes from `first_hdr_size` to the end of the comment’s NUL: `first_hdr_size` + name length + 1 + comment length + 1. At most 2,600. `0` marks the end of the archive.
`4`	…	Basic header	The fixed fields (main or local, below), any extra data, then the name and the comment, each ending with a NUL.
…	4	Basic header CRC	CRC-32 of the basic header (from offset 4, the size given at offset 2).
…	2	Extended header size	`0` when there is none (the usual case).
…	…	Extended header	Unused by ARJ 2.4x; the size and data repeat until a size of `0`.
…	4	Extended header CRC	CRC-32 of that extended header. Absent when the size is 0.

The technical note warns that older ARJ, AV.C and UNARJ.C skip only 2 bytes for an extended header’s CRC instead of 4; it did not matter because no ARJ version then wrote extended headers.

The CRC is the standard CRC-32 of ZIP (polynomial EDB88320, start FFFFFFFF, final value inverted); the note gives no parameters and points to UNARJ’s source, which uses this one.

Main header

Offsets are from the start of the header ID. first_hdr_size is 30 in ARJ 2.4x archives; later versions write 34 and add the bytes at 34 to 37.

Offset	Size	Name	Description
`4`	1	`first_hdr_size`	Size of the fixed part and extra data, counted from this byte.
`5`	1	Archiver version	Version of ARJ that made the archive.
`6`	1	Minimum version to extract
`7`	1	Host OS	See host OS.
`8`	1	ARJ flags	See main header flags.
`9`	1	Security version	`2` is current.
`10`	1	File type	Always `2` in the main header.
`11`	1	Reserved
`12`	4	Created	Date and time the archive was created, in DOS format.
`16`	4	Modified	Date and time the archive was last modified (added to the note in 1993).
`20`	4	Archive size	Used only by secured archives.
`24`	4	Security envelope position	File position of the ARJ-SECURITY envelope.
`28`	2	Filespec position	Where the file name part starts in the stored name.
`30`	2	Security envelope length
`32`	1	Encryption version	“Not used” in ARJ 2.4x; later versions record the garbling method here.
`33`	1	Last chapter	Later versions: the highest chapter number in a chapter archive. “Not used” in ARJ 2.4x.
`34`	1	Protection factor	Later versions, when `first_hdr_size` ≥ 34: the ARJ-PROTECT recovery data factor.
`35`	1	Flags 2	Later versions: second flags byte.
`36`	2	Spare	Later versions: unused.
4 + `first_hdr_size`	…	Archive name	The archive’s file name when it was made, NUL-terminated.
…	…	Comment	Archive comment, NUL-terminated.

ARJ-SECURITY, the authenticity envelope sold as an add-on, uses the fields at 9 and 20 to 31; their layout was made public with ARJ 2.39. The contents of the envelope itself are not public.

Main header flags

Bit	ARJ 2.4x name	Meaning
`01`	not used	Later versions: `GARBLED_FLAG`, the archive is password-protected.
`02`	`OLD_SECURED_FLAG`	Secured with the old method. Later versions reuse the bit as `ANSIPAGE_FLAG`: names are in the Windows ANSI code page.
`04`	`VOLUME_FLAG`	Another volume follows.
`08`	not used	Later versions: `ARJPROT_FLAG`, the archive carries ARJ-PROTECT recovery data.
`10`	`PATHSYM_FLAG`	The archive name was translated: `\` became `/`.
`20`	`BACKUP_FLAG`	A backup-type archive.
`40`	`SECURED_FLAG`	The archive has an ARJ-SECURITY envelope.
`80`	–	Later versions: `ALTNAME_FLAG`. Its exact meaning is not in any public description found.

Local file header

first_hdr_size is 30, or 34 when EXTFILE_FLAG is set; later versions write 46 to add the access and creation times.

Offset	Size	Name	Description
`4`	1	`first_hdr_size`	As in the main header.
`5`	1	Archiver version
`6`	1	Minimum version to extract
`7`	1	Host OS	See host OS.
`8`	1	ARJ flags	See local header flags.
`9`	1	Method	See methods.
`10`	1	File type	See file types.
`11`	1	Reserved	Later versions: the password modifier, a value mixed into the garbling key.
`12`	4	Modified	Date and time the file was last modified, in DOS format.
`16`	4	Compressed size	Bytes of data after the header.
`20`	4	Original size	Size when extracted. For text-mode files this can differ from the size on disk, because line endings are converted.
`24`	4	Original CRC	CRC-32 of the extracted data.
`28`	2	Filespec position	Offset of the file name part in the stored path.
`30`	2	File access mode	File attributes or mode of the host system (DOS attributes on DOS, `st_mode` bits on Unix).
`32`	1	First chapter	“Host data, currently unused” in ARJ 2.4x. Later versions: the first chapter the file belongs to.
`33`	1	Last chapter	Later versions: the last chapter the file belongs to.
`34`	4	Extended file position	Only with `EXTFILE_FLAG`: where this part of a split file starts in the whole file.
`38`	4	Accessed	Later versions, when `first_hdr_size` ≥ 46: last access time, DOS format.
`42`	4	Created	As above: creation time.
`46`	4	Original size, all volumes	As above: the whole file’s size even when this part is one volume of it.
4 + `first_hdr_size`	…	File name	Path and name, NUL-terminated.
…	…	Comment	File comment, NUL-terminated.

The compressed data follows the frame’s last CRC (or the zero extended header size).

Local header flags

Bit	Name	Meaning
`01`	`GARBLED_FLAG`	The file is password-protected (garbled).
`02`	not used
`04`	`VOLUME_FLAG`	The file continues in the next volume.
`08`	`EXTFILE_FLAG`	This is a continued part: the extended file position field is present.
`10`	`PATHSYM_FLAG`	The name was translated: `\` became `/`.
`20`	`BACKUP_FLAG`	The file is marked as a backup.
`40`, `80`	–	Not defined in the note.

Code values

Host OS

Value	System
`0`	MS-DOS
`1`	PRIMOS
`2`	Unix
`3`	Amiga
`4`	Mac OS
`5`	OS/2
`6`	Apple IIGS
`7`	Atari ST
`8`	NeXT
`9`	VAX VMS
`10`	Windows 95 (later versions)
`11`	Win32 (later versions)

Methods

Value	Method
`0`	Stored
`1`	Compressed, most (best compression)
`2`	Compressed
`3`	Compressed, faster
`4`	Compressed, fastest
`8`	No data, no CRC (later versions; used for entries such as chapter labels)
`9`	No data (later versions)

Methods 1 to 3 are the same LZ77 and static Huffman scheme, the one UNARJ decodes, with a 26,624-byte window; they differ only in how hard the compressor searches. Method 4 is a simpler and faster scheme with no Huffman tables: match lengths and positions are written as variable-length bit codes. UNARJ’s source is the public description of both decoders.

File types

Value	Type
`0`	Binary
`1`	7-bit text (stored with line endings converted)
`2`	Main header (later versions also: comment header)
`3`	Directory
`4`	Volume label
`5`	Chapter label (later versions)

Date and time

A 32-bit MS-DOS date and time, local time, with 2-second resolution:

Bits	Meaning
31 to 25	Year − 1980
24 to 21	Month, 1 to 12
20 to 16	Day, 1 to 31
15 to 11	Hour, 0 to 23
10 to 5	Minute, 0 to 59
4 to 0	Second ÷ 2

Finding the archive

A self-extracting ARJ is an MS-DOS program with the archive after it, and the main header need not follow the program directly. The note’s find_header() routine searches for it:

Find the bytes 60 EA.
Read the next 2 bytes as the basic header size. If it is over 2,600, move one byte past the ID and search again.
Read that many bytes and their CRC-32. If the CRC does not match, move one byte past the ID and search again.

The archive ends with a header whose basic header size is 0: the four bytes 60 EA 00 00. Split archives continue in .a01, .a02 and so on, each with its own main header.

.ARJ File Format – Headers and codes