Skip to content
Switch branches/tags
Go to file
Cannot retrieve contributors at this time
PasswordSafe database format description version 4.01
Copyright (c) 2013-2021 Rony Shapiro <>.
All rights reserved. Use of the code is allowed under the Artistic
License terms, as specified in the LICENSE file distributed with this
code, or available from
1. Introduction: This document defines a file format for the secure
storage of passwords and related data. The format is designed
according to current cryptographic best practices, and is believed to
be secure, in the sense that without knowledge of the master
passphrase, only a brute-force attack or a flaw in the underlying
cryptographic algorithms will result in unauthorized access to the
1.1 Design Goals: The PasswordSafe database format is designed to be
secure, extensible and platform-independent.
1.2 History: This specification is an evolution of previous
formats. The main differences between version 4 of the format and
previous versions are:
1.2.1 Support for Multiple Independent Master Passwords
The ability to have > 1 master password for a given database is less
important for the individual user, but is a repeated request in
corporate settings. This would allow 'escrow' services, protect
against a single user becoming unavailable, and allow splitting of
passwords into groups with different users for each group by using
multiple databases.
1.2.2 Support Multiple Encryption Algorithms
In addition to Twofish (which will remain the default), AES256 will be
supported. Other algorithms may be added in the future.
1.2.3 Support Attachments
There's demand, mainly on the mobile ports, for the ability to store
attachments with password entries. The main use-case is for photos
taken by mobile phones, e.g., passports, credit cards, etc. Therefore,
it seems reasonable to allow adding, deleting and viewing images for
As the size of an attachment can be large relative to the database, we
would like to avoid reading and writing it unless and until
necessary. Note that this is a break with the current paradigm of
reading the entire database into memory once upon opening.
1.2.4 Other Changes
This Section summarizes minor changes from the previous format
version, as is informative, not normative. Hex to Binary
Several numeric fields in are encoded in V3 as ASCII hex data, which
is inefficient and arguably inelegant. All such fields have been
changed to little-endian integer representations. Time Representation
Time is uniformly represented as 40 bit unsigned little-endian
integers, in seconds from 1/1/1970, UTC. V3 representation is 32 bits. Fix Minor Weakness
The only known weakness in the V3 format is that the HMAC
integrity check does not currently cover the field type and field
length values. This has been fixed.
1.3 Backwards Compatibility:
This format is NOT compatible with previous (major) versions. Note,
however, that since the data stored in previous versions is a proper
subset of the data described here, implementers may read a database
written in an older version and store the result in the format
described here.
2. Format: A V4 format PasswordSafe is structured as follows:
2.1 Nonce is a 256 bit random value, used (a) to verify that the file
has not changed since last read (although this is not foolproof,
it's the best we can do across file sharers), and (b) to help detect
the end of the key blocks, as described below.
2.2 KB1..KBn represent Key Blocks, where each key block contains the
encryption and integrity verification keys, wrapped as follows:
2.2.1 SALT is a 256 bit random value, generated at Key Block creation
2.2.2 ITER is the number of iterations on the hash function to
calculate Pi', stored as a 32 bit little-endian value. This value is
stored here in order to future-proof the file format against increases
in processing power.
2.2.3 Pi' is the "stretched key" for Key Block i, generated from the
user's passphrase (UTF-8 encoded) and the SALT, as defined by the
hash-function-based key stretching algorithm in [PBKDF2], with SHA-256
[SHA256] as the hash function, and ITER iterations (at least
2048). The intention is that Key Block i will be generated with user
i's passphrase. All keyblocks protect the same K and L values (see
2.2.4 KW is the Key Wrap algorithm defined in [RFC3394], using the
user-specified encryption algorithm (TwoFish, AES,...).
2.2.5 KW(Pi',K) is the wrapped random encryption key K that is used to
encrypt header HDR and records R1..Rn. As K is 256 bits, KW(Pi',K) is
320 bits (40 bytes).
2.2.6 KW(Pi',L) is the wrapped random key L that is used to
calculate the HMAC (keyed-hash message authentication code) of the
data. See description of EOF field below for more details. As L is 256
bits, KW(Pi',K) is 320 bits (40 bytes).
Implementation Note: K and L must NOT be related.
2.3 endKB consists of two parts: The first is a 256 bit indicator of
the end of the Key Blocks. This is the SHA-256 value of the Nonce
described in 2.1. The idea is not to give a plaintext that's too easy
to scan, but otherwise this value has no cryptographic utility. The
second 256 bit value is HMAC(SHA256, L, KB1..KBn|SHA256(Nonce)). This
serves as an integrity check on the Key Blocks.
2.4 IV is the 128-bit random Initial Value for CBC mode.
2.5 All following records are encrypted using the user-selected block
cipher in CBC mode, with K as the encryption key.
2.5.1 HDR: The database header. The header consists of one or more typed
fields (as defined in section 3.2), beginning with the Version type
field, and terminated by the 'END' type field. The version number
and END fields are mandatory. Aside from these two fields, no order is
assumed on the field types.
2.5.2 R1..Rn: The actual database records. Each record consists of one or
more typed fields (as defined in Section 3.3), terminated by the 'END' type
field. The first field of a record is UUID field, of one of several
possible UUID types. The specific type defines the record type, and the
contents of the field is the UUID of the record.
Currently the following record types are defined: An "Entry" record. This represents a password entry. Aside
from the UUID field, the Title, and Password fields are mandatory. An "Alias" record. This record is identical to an Entry
record, except that instead of a Password field, it has a Reference to
the "base" Entry record, from which it takes the password value at
runtime. (The use-case for this is a single-sign-on site, where the
user has access to several systems that share the same password).
The mandatory fields for this record are:
- AliasUUID: The UUID of the record
- Title
- BaseUUID - the UUID of the referred-to Entry record
- END A "Shortcut" record. This record also refers to a "base" Entry
record, but, unlike an alias, takes all of the field values from the
base Entry except for title and (optionally) group and user. The
fields for this record are, therefore:
- ShortcutUUID (mandatory)
- Title (mandatory)
- Group (optional)
- User (optional)
- BaseUUID - the UUID of the referred-to Entry record
- END An "Attachment" record. This record is used to contain
attachments, e.g., jpeg, tiff, or other media types. This record is
defined to be "skipped over" while reading, and to be copied without
the need to decrypt and re-encrypt. See Section 3.4 for details. For all records, all non-mandatory fields may either be absent
or have zero length. When a field is absent or zero-length, its
default value shall be used. Aside from first UUID field and the
terminating 'END' field, no order is assumed on the field types unless
otherwise specified.
2.6 HMAC: The 256-bit keyed-hash MAC, as described in RFC2104, with SHA-
256 as the underlying hash function. The value is calculated over the
plaintext values of the field type, the field length and the data
stored in the field, over all fields (starting from the version number
in the header, ending with the last field of the last record). The
exception is the content field (of the attachment record), which is
handled separately as described in Section 3.4.2. The key L is used as
the hash key value.
3. Fields: Data in PasswordSafe is stored in typed fields. Each field
consists of one or more blocks. The blocks are the blocks of the underlying
encryption algorithm - 16 bytes long for Twofish. The first block contains
the field length in the first 4 bytes (little-endian), followed by a one-
byte type identifier. The rest of the block contains up to 11 bytes of
record data. If the record has less than 11 bytes of data, the extra bytes
are filled with random values. The type of a field also defines the data
3.1 Data representations
3.1.1 UUID
The UUID data type is 16 bytes long, as defined in [RFC4122]. Microsoft
Windows has functions for this, and the RFC has a sample
3.1.2 Text
Text is represented in UTF-8 encoding (as defined in [RFC3629]), with
no byte order marker (BOM) and no end-of-string mark (e.g., null
byte). Note that the latter isn't necessary since the length of the
field is provided explicitly. ALL fields described as "text" are
UTF-8 encoded unless explicitly stated otherwise.
3.1.3 Time
Timestamps are stored as 40 bit, little endian, unsigned integers,
representing the number of seconds since Midnight, January 1, 1970, GMT.
(This is equivalent to the time_t type on Windows and POSIX. On the
Macintosh, the value needs to be adjusted by the constant value 2082844800
to account for the different epoch of its time_t type.)
3.2 Field types for the PasswordSafe database header:
Name Value Type Implemented Comments
Version 0x00 2 bytes Y [1]
UUID 0x01 UUID Y [2]
Non-default preferences 0x02 Text Y [3]
Tree Display Status 0x03 Text Y [4]
Timestamp of last save 0x04 time_t Y [5]
Who performed last save 0x05 Text Y [OBSOLETE 6]
What performed last save 0x06 Text Y [7]
Last saved by user 0x07 Text Y [8]
Last saved on host 0x08 Text Y [9]
Database Name 0x09 Text Y [10]
Database Description 0x0a Text Y [11]
Database Filters 0x0b Text Y [12]
Reserved 0x0c - [13]
Reserved 0x0d - [13]
Reserved 0x0e - [13]
Recently Used Entries 0x0f Text [14]
Named Password Policies 0x10 Text [15]
Empty Groups 0x11 Text [16]
Yubico 0x12 Text [13]
Timestamp of last master
password change 0x13 time_t Y [5]
End of Entry 0xff [empty] Y [17]
[1] The version number of the database format. For this version, the value
is 0x0400 (stored in little-endian format, that is, 0x00, 0x04).
[2] A universally unique database identifier is needed in order to
synchronize databases across different machine. Representation is
described in Section 3.1.1.
[3] Non-default preferences are encoded in a string as follows: The
string is of the form "X nn vv X nn vv..." Where X=[BIS] for boolean,
integer and string respectively, nn is the numeric value of the enum,
and vv is the value, {1 or 0} for bool, unsigned integer for int, and
a delimited string for String. Only non-default values are stored. See
PWSprefs.cpp for more details. Note: normally strings are delimited
by the doublequote character. However, if this character is in the
string value, an arbitrary character will be chosen to delimit the
[4] If requested to be saved, this is a string of 1s and 0s indicating the
expanded state of the tree display when the database was saved. This can
be applied at database open time, if the user wishes, so that the tree is
displayed as it was. Alternatively, it can be ignored and the tree
displayed completely expanded or collapsed. Note that the mapping of
the string to the display state is implementation-specific.
[5] Representation is as described in Section 3.1.3. Note that prior
to PasswordSafe 3.09, this field was mistakenly represented as an
eight-byte hexadecimal ASCII string. Implementations SHOULD attempt to
parse 8-byte long timestamps as a hexadecimal ASCII string
representation of the timestamp value.
[6] This field was deprecated in the V3 format, and should not be read
or written in a V4 database. If the field appears while importing an
old V3 database, the application should convert it to the fields
replacing it as described in the V3 file format.
[7] Free form text giving the application that saved the database.
For example, the Windows PasswordSafe application will use the text
"Password Safe", where nn and mm are the major and minor
version numbers. The major version will contain only the significant
digits whereas the minor version will be padded to the left with
zeroes e.g. "Password Safe V3.02".
[8] Text containing the username (e.g., login, userid, etc.) of the
user who last saved the database, as determined by the appropriate
operating-system dependent function.
[9] Text containing the hostname (e.g., machine name, hostid, etc.) of the
machine on which the database was last saved, as determined by the
appropriate operating-system dependent function.
[10] Database name. A logical name for a database which can be used by
applications in place of the possibly lengthy filepath notion. Note
that this field SHOULD be limited to what can be displayed in a single
[11] Database Description. A purely informative description concerning
the purpose or other practical use of the database.
[12] Specific filters for this database. This is the text equivalent to
the XML export of the filters as defined by the filter schema. The text
'image' has no 'print formatting' e.g. tabs and carriage return/line feeds,
since XML processing does not require this.
[13] Values marked 'Reserved' are know to have been used by customized
versions of PasswordSafe. To ensure compatability between versions,
use of these values should be avoided.
[14] A list of the UUIDs (16 byte binary representation) of the
recently used entries, prefixed by a one byte representation of the
number of these entries. The entries are stored in the order of last
access, with the first entry being the most recent entry accessed.
[15] Database-specific Password Policies. The format is:
N = 1 byte specifying the number of password policies in field
(max. 255).
Each entry is of the form:
L = 1 byte specifying length of the policy name in bytes = The policy name (maximum 255 bytes, UTF-8, non-null terminated)
ff = 2 bytes in little-endian order representing the following flags
UseLowercase 0x8000
UseUppercase 0x4000
UseDigits 0x2000
UseSymbols 0x1000
UseHexDigits 0x0800 (if set, then no other flags can be set)
UseEasyVision 0x0400
MakePronounceable 0x0200
Unused 0x01ff
nn = 2 bytes in little-endian order password total length
ll = 2 bytes in little-endian order minimum number of lowercase
characters in generated password
uu = 2 bytes in little-endian order minimum number of uppercase
characters in generated password
dd = 2 bytes in little-endian order minimum number of digit
characters in generated password
ss = 2 bytes in little-endian order minimum number of symbol
characters in generated password
M = 1 byte giving the length in bytes of allowed special symbols,
or zero to specify the database default special symbol set.
SSS...SSS = List of allowed symbols in this policy (maximum 255 bytes)
[16] This fields contains the name of an empty group that cannot be
constructed from entries within the database. Unlike other header
fields, this field can appear multiple times.
[17] An explicit end of entry field is useful for supporting new fields
without breaking backwards compatability.
3.3 Field types for database Records:
Name Value Type Implemented Comments
UUID 0x01 UUID Y [1]
Group 0x02 Text Y [2]
Title 0x03 Text Y
Username 0x04 Text Y
Notes 0x05 Text Y
Password 0x06 Text Y
Creation Time 0x07 time_t Y [5]
Password Modification Time 0x08 time_t Y [5]
Last Access Time 0x09 time_t Y [5,6]
Password Expiry Time 0x0a time_t Y [5,7]
*RESERVED* 0x0b 4 bytes - [8]
Last Modification Time 0x0c time_t Y [5,9]
URL 0x0d Text Y [10]
Autotype 0x0e Text Y [11]
Password History 0x0f Text Y [12]
Password Policy 0x10 Text Y [13]
Password Expiry Interval 0x11 4 bytes Y [14]
Run Command 0x12 Text Y
Double-Click Action 0x13 2 bytes Y [15]
EMail address 0x14 Text Y [16]
Protected Entry 0x15 1 byte Y [17]
Own symbols for password 0x16 Text Y [18]
Shift Double-Click Action 0x17 2 bytes Y [15]
Password Policy Name 0x18 Text Y [19]
Entry keyboard shortcut 0x19 4 bytes Y [20]
AttRef 0x1a UUID Y [21]
Two-Factor Key 0x1b Binary N [22]
Credit Card Number 0x1c Text N [23]
Credit Card Expiration 0x1d Text N [23]
Credit Card Verif. Value 0x1e Text N [23]
Credit Card PIN 0x1f Text N [23]
QR Code 0x20 Text N [24]
BaseUUID 0x41 UUID Y [25]
AliasUUID 0x42 UUID Y
ShortcutUUID 0x43 UUID Y
Unknown (testing) 0xdf - N [26]
Implementation-specific 0xe0-0xfe - N [27]
End of Entry 0xff [empty] Y [28]
[1] UUID of a password entry record. Representation is described in
Section 3.1.1.
[2] The "Group" supports displaying the entries in a tree-like manner.
Groups can be hierarchical, with elements separated by a period, supporting
groups such as " cards.Visa", "
cards.Mastercard", access", etc. Dots entered by the user
should be "escaped" by the application.
[5] Representation is as described in Section 3.1.3.
[6] This will be updated whenever any part of this entry is accessed
i.e., to copy its username, password or notes to the clipboard; to
perform autotype or to browse to url.
[7] This will allow the user to enter an expiry date for an entry. The
application can then prompt the user about passwords that need to be
changed. A value of zero means "forever".
[8] Although earmarked for Password Policy, the coding in versions prior
to V3.12 does not correctly handle the presence of this field. For this
reason, this value cannot be used for any future V3 field without causing
a potential issue when a user opens a V3.12 or later database with program
version V3.11 or earlier. See note [14].
[9] This is the time that any field of the record was modified, useful for
merging databases.
[10] The URL will be passed to the shell when the user chooses the "Browse
to" action for this entry. In version 2 of the format, this was extracted
from the Notes field. By placing it in a separate field, we are no longer
restricted to a URL - any action that may be executed by the shell may be
specified here.
[11] The text to be 'typed' by PasswordSafe upon the "Perform Autotype"
action maybe specified here. If unspecified, the default value of
'username, tab, password, tab, enter' is used. In version 2 of the format,
this was extracted from the Notes field. Several codes are recognized here,
e.g, '\p' is replaced by the record's password. See the user documentation
for the complete list of codes. The replacement is done by the application
at runtime, and is not stored in the database.
[12] Password History is an optional record. If it exists, it stores the
creation times and values of the last few passwords used in the current
entry, in the following format:
f = {0,1} if password history is off(0) / on(1)
mm = 2 hexadecimal digits max size of history list (i.e. max = 255)
nn = 2 hexadecimal digits current size of history list
T = Time password was set (time_t written out in %08x)
L = 4 hexadecimal digit password length (in TCHAR)
P = Password
No history being kept for a record can be represented either by the lack of
the PWH field (preferred), or by a header of _T("00000"):
flag = 0, max = 00, num = 00
Note that 0aabb, where bb <= aa, is possible if password history was enabled
in the past and has then been disabled but the history hasn't been cleared.
[13] This field allows a specific Password Policy per entry. This field
is mutually exclusive with the policy name field [0x18]. The format is:
ffff = 4 hexadecimal digits representing the following flags
UseLowercase 0x8000
UseUppercase 0x4000
UseDigits 0x2000
UseSymbols 0x1000
UseHexDigits 0x0800 (if set, then no other flags can be set)
UseEasyVision 0x0400
MakePronounceable 0x0200
Unused 0x01ff
nnn = 3 hexadecimal digits password total length
lll = 3 hexadecimal digits password minimum number of lowercase characters
uuu = 3 hexadecimal digits password minimum number of uppercase characters
ddd = 3 hexadecimal digits password minimum number of digit characters
sss = 3 hexadecimal digits password minimum number of symbol characters
[14] Password Expiry Interval, in days, before this password expires. Once set,
this value is used when the password is first generated and thereafter whenever
the password is changed, until this value is unset. Valid values are 1-3650
corresponding to up to approximately 10 years. A value of zero is equivalent to
this field not being set. Value is stored in little-endian format.
[15] A two byte field contain the value of the Double-Click Action and Shift
Double-Click Action'preference value' (0xff means use the current
Application default):
Current 'preference values' are:
CopyPassword 0
ViewEdit 1
AutoType 2
Browse 3
CopyNotes 4
CopyUsername 5
CopyPasswordMinimize 6
BrowsePlus 7
Run Command 8
Send email 9
[16] Separate Email address field as per RFC 2368 (without the 'mailto:'
prefix. This field was introduced in version 0x0306 (PasswordSafe V3.19).
[17] Entry is protected, i.e., the entry cannot be changed or deleted
while this field is set. This field was introduced in version 0x0308
(PasswordSafe V3.25). This a single byte. An absent field or a zero
valued field means that the entry is not protected. Any non-zero value
means that the entry is protected.
[18] Each entry can now specify its own set of allowed special symbols for
password generation. This overrides the default set and any database specific
set. This field is mutually exclusive with the policy name field
[0x18]. This was introduced in version 0x0309 (PasswordSafe V3.26).
[19] Each entry can now specify the name of a Password Policy saved in
the database header for password generation. This field is mutually
exclusive with the specific policy field [0x10] and with the Own
symbols for password field [0x16]. This was introduced in version
0x030A (PasswordSafe V3.28).
[20] Entry keyboard shortcut. Format is:
Bytes 0-1: Virtual KeyCode for the character (Windows only uses byte 0)
Byte 2 : Keyboard Modifiers - a bitwise OR of any valid combination of:
PWS_HOTKEYF_META 0x10 (not supported by Windows)
PWS_HOTKEYF_WIN 0x20 (not supported by Windows)
PWS_HOTKEYF_CMD 0x40 (not supported by Windows)
Byte 3 : zero
This was introduced in version 0x030D (PasswordSafe V3.30).
[21] The UUID of an attachment record associated with an entry. An
entry record can have zero or more such fields. Each attachment field
in a given password entry record must have a different value, but
different entries can refer to the same attachment uuid.
[22] This is the shared secret for sites using Time-Based One-Time
Password Algorithm (per RFC6238) such as Google Authenticator. At
least 10 bytes.
[23] Credit card specific fields. All values are UTF-8 encoded:
- Number should consist of digits and spaces.
- Expiration should be MM/YY, where MM is 01-12, and YY 00-99.
- CVV (CVV2) is three or four digits.
- PIN is four to twelve digits long (ISO-9564)
[24] UTF-8 encoded text used for QR code generation. Content will be converted
to QR code without any further encoding.
[25] UUID of a reference record. Representation is described in Section
[26] Reserved for testing forward compatability.
[27] Reserved for application-specific use. See 4.2.2.
[28] An explicit end of entry field is useful for supporting new fields
without breaking backwards compatability.
3.4 Attachments
3.4.1 Attachment record fields
An Attachment record has the following fields:
Name Value Type Mandatory Comments
AttUUID 0x60 UUID Y [1]
Title 0x61 Text N [2]
Creation Time 0x62 time_t Y [3]
MediaType 0x63 Text Y [4]
FileName 0x64 Text N [5]
FilePath 0x65 Text N [6]
FileCreationTime 0x66 time_t N [7]
FileModificationTime 0x67 time_t N [7]
FileAccessTime 0x68 time_t N [7]
AttEK 0x70 32 bytes Y [8]
AttAK 0x71 32 bytes Y [9]
ATTIV 0x71 binary Y [10]
Content 0x73 4 byte+binary Y [11]
ContentHMAC 0x74 32 bytes Y [12]
End of Entry 0xff [empty] Y
[1] This is the UUID of the attachment record, which is referred to by
password entries' AttRef field. Representation is described in Section
3.1.1. This is the first field in the record, and identifies it as an
attachment record.
[2] The title is a user-friendly name for the attachment.
[3] The time when the attachment was added. Representation is as
described in Section 3.1.3.
[4] MIME type - text string describing Media Type, per
[5] Original file name - kept mainly for the extension, to support
applications that require this for proper viewing.
[6] Path component of file. May be relative or absolute (or absent).
[7] Times associated with the attachment file at the time of attachment.
Application may try to restore these when exporting the attachment.
Representation is as described in Section 3.1.3.
[8] Attachment encryption key: This is a 256 bit random key encrypting the
[9] Attachment authentication key: This is a 256 bit random key used to
verify the integrity of the content.
[10] The initialization vector (IV) used in the CBC encryption of the
attachment. Size is the blocksize of the underlying block cipher.
[11] The binary representation of the attachment. Note that this is limited
to 4GB (2^32 bytes), which should suffice.
The first block of this field is encrypted under the database's
encryption key. This is so the field's type and length can be determined
without relying on the implicit order of the fields. The length of
this field is 4, and its contents is 4 byte little-endian value of the
length of the actual content, padded as usual with random data. The
actual content is in the following blocks, that are encrypted under
the AttEK in CBC mode with the value of ATTIV as the IV, using the
same algorithm as the rest of the database. Padding of content is done
with random data, as needed.
[12] The content HMAC is calculated over the same plaintext data that
is protected using the AttEK, including padding. HMAC calculation uses
AttAK and SHA256.
This field is encrypted with K (see 2.2.5). The first block of the
Content field is used as the IV, so as to continue the database CBC
stream. This field must immediately follow the Content field.
3.4.2 Changes to database HMAC calculation The HMAC of the database file will not be calculated over the data
that is covered by content field HMAC. This enables saving databases
without the need to recalculate the HMAC of contents. The database HMAC will cover the attachment records' other fields,
specifically the content HMAC key and the HMAC calculated under that key.
The database HMAC will also cover the first block of the attachment content
fields, as these are not covered by content HMAC.
3.4.3 Implementation note
When reading the database, it is assumed that the offsets of the attachment
records will be noted for rapid access to the attachments when
required. Care must be taken to ensure that the database was not modified
between when the offsets were recorded and when they are actually
4. Extensibility
4.1 Forward compatability: Implementations of this format SHOULD NOT
discard or report an error when encountering a field of an unknown
type. Rather, the field(s) type and data should be read, and preserved
when the database is saved.
4.2 Field type identifiers: This document specifies the field type
identifiers for the current version of the format. Compliant
implementations MUST support the mandatory fields, and SHOULD support
the other fields described herein. Future versions of the format may
specify other type identifiers.
4.2.1 Application-unique type identifiers: The type identifiers
0xc0-0xdf are available for application developers on a first-come
first-serve basis. Application developers interested in reserving a
type identifier for their application should contact the maintainer of
this document (Currently the PasswordSafe project administrator at
4.2.2 Application-specific type identifiers: The type identifiers
0xe0-0xfe are reserved for implementation-specific purposes, and will
NOT be specified in this or future versions of the format
4.2.3 All unassigned identifiers except as listed in the previous two
subsections are reserved, and should not be used by other
implementations of this format specification in the interest of
5. References:
End of Format description.