1) Extensible (can incorporate user-defined information). 2) Meaningful tag ID's (hint to meaning of unknown information). 3) Sequential read/write ability (streamable). 4) Hierarchical information structure. 5) Easy to implement reader/writer/editor. 6) Order of information well defined. 7) Large data lengths supported: >64kB (+5) and >4GB (+5). 8) Localized text strings. 9) Multiple documents in a single file. 10) Compact format doesn't squander disk space or bandwidth. 11) Compressed meta information supported. 12) Relocatable data elements (ie. no fixed offsets). 13) Binary meta information (+7) with variable byte order (+3). 14) Mandatory tags not required (an unnecessary complication). 15) Append information to end of file without editing. Feature number Total Format 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Score ------ --------------------------------------------- ----- MIE 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 150 PDF 10 10 0 10 0 0 10 0 10 10 10 0 7 10 10 97 PNG 10 10 10 0 8 0 5 10 0 10 10 10 0 10 0 93 XMP 10 10 10 10 2 0 10 10 10 0 0 10 0 10 0 92 AIFF 0 5 10 10 10 0 5 0 0 10 0 10 7 10 0 77 RIFF 0 5 10 10 10 0 5 0 0 10 0 10 7 10 0 77 JPEG 10 0 10 0 10 0 0 0 0 10 0 10 7 10 0 67 EPS 10 10 10 0 0 0 10 0 10 0 0 5 0 10 0 65 CIFF 0 0 0 10 10 0 5 0 0 10 0 10 10 10 0 65 TIFF 0 0 0 10 5 10 5 0 10 10 0 0 10 0 0 60 EXIF 0 0 0 10 5 10 0 0 0 10 0 0 10 0 0 45 IPTC 0 0 10 0 8 0 0 0 0 10 0 10 7 0 0 45
By design, MIE ranks highest by a significant margin. Other formats with reasonable scores are PDF, PNG and XMP, but each has significant weak points. What may be surprising is that TIFF, EXIF and IPTC rank so low.
As well as scoring high in all these features, the MIE format has the unique ability to encapsulate any other type of file, and provides a non-invasive method of adding meta information to a file. The meta information is logically separated from the original file data, which is extremely important because meta information is routinely lost when files are edited.
Also, the MIE format supports multiple files by simple concatenation, enabling all kinds of wonderful features such as linear databases, edit histories or non-intrusive file updates. This ability can also be leveraged to allow MIE-format trailers to be added to some other file types.
Byte Number: 0 1 2 3 4 5 6 7 C Characters: ~ \x10 \x04 ? 0 M I E or ~ \x18 \x04 ? 0 M I E Hexadecimal: 7e 10 04 ? 30 4d 49 45 or 7e 18 04 ? 30 4d 49 45 Decimal: 126 16 4 ? 48 77 73 69 or 126 24 4 ? 48 77 73 69
Note that byte 1 may have one of the two possible values (0x10 or 0x18), and byte 3 may have any value (0x00 to 0xff).
1 byte SyncByte = 0x7e (decimal 126, character '~') 1 byte FormatCode (see below) 1 byte TagLength (T) 1 byte DataLength (gives D if DataLength < 253) T bytes TagName (T given by TagLength) 2 bytes DataLength2 [exists only if DataLength == 255 (0xff)] 4 bytes DataLength4 [exists only if DataLength == 254 (0xfe)] 8 bytes DataLength8 [exists only if DataLength == 253 (0xfd)] D bytes DataBlock (D given by DataLength)
The minimum element length is 4 bytes (for a group terminator). The maximum DataBlock size is 2^64-1 bytes. TagLength and DataLength are unsigned integers, and the byte ordering for multi-byte DataLength fields is specified by the containing MIE group element. The SyncByte is byte aligned, so no padding is added to align on an N-byte boundary.
The format code is a bitmask that defines the format of the data:
7654 3210 ++++ ---- FormatType ---- +--- TypeModifier ---- -+-- Compressed ---- --++ FormatSize
0x00 - other (or unknown) data 0x10 - MIE group 0x20 - text string 0x30 - list of null-separated text strings 0x40 - integer 0x50 - rational 0x60 - fixed point 0x70 - floating point 0x80 - free space
0x08 - other data sensitive to MIE group byte order 0x18 - MIE group with little-endian byte ordering 0x28 - UTF encoded text string 0x38 - UTF encoded text string list 0x48 - signed integer 0x58 - signed rational (denominator is always unsigned) 0x68 - signed fixed-point
0x00 - 8 bits (1 byte) 0x01 - 16 bits (2 bytes) 0x02 - 32 bits (4 bytes) 0x03 - 64 bits (8 bytes)
The number of bytes in a single value for this format is given by 2**FormatSize (or 1 << FormatSize). The number of values is the data length divided by this number of bytes. It is an error if the data length is not an even multiple of the format size in bytes.
The following is a list of all currently defined MIE FormatCode values for uncompressed data (add 0x04 to each value for compressed data):
0x00 - other data (insensitive to MIE group byte order) (1) 0x01 - other 16-bit data (may be byte swapped) 0x02 - other 32-bit data (may be byte swapped) 0x03 - other 64-bit data (may be byte swapped) 0x08 - other data (sensitive to MIE group byte order) (1) 0x10 - MIE group with big-endian values (1) 0x18 - MIE group with little-endian values (1) 0x20 - ASCII (ISO 8859-1) string (2,3) 0x28 - UTF-8 string (2,3,4) 0x29 - UTF-16 string (2,3,4) 0x2a - UTF-32 string (2,3,4) 0x30 - ASCII (ISO 8859-1) string list (3,5) 0x38 - UTF-8 string list (3,4,5) 0x39 - UTF-16 string list (3,4,5) 0x3a - UTF-32 string list (3,4,5) 0x40 - unsigned 8-bit integer 0x41 - unsigned 16-bit integer 0x42 - unsigned 32-bit integer 0x43 - unsigned 64-bit integer (6) 0x48 - signed 8-bit integer 0x49 - signed 16-bit integer 0x4a - signed 32-bit integer 0x4b - signed 64-bit integer (6) 0x52 - unsigned 32-bit rational (16-bit numerator then denominator) (7) 0x53 - unsigned 64-bit rational (32-bit numerator then denominator) (7) 0x5a - signed 32-bit rational (denominator is unsigned) (7) 0x5b - signed 64-bit rational (denominator is unsigned) (7) 0x61 - unsigned 16-bit fixed-point (high 8 bits is integer part) (8) 0x62 - unsigned 32-bit fixed-point (high 16 bits is integer part) (8) 0x69 - signed 16-bit fixed-point (high 8 bits is signed integer) (8) 0x6a - signed 32-bit fixed-point (high 16 bits is signed integer) (8) 0x72 - 32-bit IEEE float (not recommended for portability reasons) 0x73 - 64-bit IEEE double (not recommended for portability reasons) (6) 0x80 - free space (value data does not contain useful information)
FormatSize Change in Byte Sequence -------------- ----------------------------------- 0x00 (8 bits) 0 1 2 3 4 5 6 7 --> 0 1 2 3 4 5 6 7 (no change) 0x01 (16 bits) 0 1 2 3 4 5 6 7 --> 1 0 3 2 5 4 7 6 0x02 (32 bits) 0 1 2 3 4 5 6 7 --> 3 2 1 0 7 6 5 4 0x03 (64 bits) 0 1 2 3 4 5 6 7 --> 7 6 5 4 3 2 1 0
Rational values consist of two integers, so they are swapped as the next lower FormatSize. For example, a 32-bit rational (FormatSize 0x02, and FormatCode 0x52 or 0x5a) is swapped as two 16-bit values (ie. as if it had FormatSize 0x01).
16-bit fixed value = 16-bit integer value / 256.0 32-bit fixed value = 32-bit integer value / 65536.0
Gives the length of the TagName string. Any value between 0 and 255 is valid, but the TagLength of 0 is valid only for the MIE group terminator.
DataLength is an unsigned byte that gives the number of bytes in the data block. A value between 0 and 252 gives the data length directly, and numbers from 253 to 255 are reserved for extended DataLength codes. Codes of 255, 254 and 253 indicate that the element contains an additional 2, 4 or 8 byte unsigned integer representing the data length.
0-252 - length of data block 255 (0xff) - use DataLength2 254 (0xfe) - use DataLength4 253 (0xfd) - use DataLength8
A DataLength of zero is valid for any element except a compressed MIE group. A zero DataLength for an uncompressed MIE group indicates that the group length is unknown. For other elements, a zero length indicates there is no associated data. A terminator element must have a DataLength of 0, 6 or 10, and may not use an extended DataLength.
The TagName string is 0 to 255 bytes long, and is composed of the ASCII characters A-Z, a-z, 0-9 and underline ('_'). Also, a dash ('-') is used to separate the language/country code in the TagName of a localized text string, and a units string (possibly containing other ASCII characters) may be appear in brackets at the end of the TagName. The TagName string is NOT null terminated. A MIE element with a tag string of zero length is reserved for the group terminator.
MIE elements are sorted alphabetically by TagName within each group. Multiple elements with the same TagName are allowed, even within the same group.
TagNames should be meaningful. Case is significant. Words should be lowercase with an uppercase first character, and acronyms should be all upper case. The underline (``_'') is provided to allow separation of two acronyms or two numbers, but it shouldn't be used unless necessary. No separation is necessary between an acronym and a word (eg. ``ISOSetting'').
All TagNames should start with an uppercase letter. An exception to this rule allows tags to begin with a digit (0-9) if they must come before other tags in the sort order, or a lowercase letter (a-z) if they must come after. For instance, the '0Type' element begins with a digit so it comes before, and the 'data' element begins with a lowercase letter so that it comes after meta information tags in the main ``0MIE'' group.
Tag names for localized text strings have an 6-character suffix with the following format: The first character is a dash ('-'), followed by a 2-character lower case ISO 639-1 language code, then an underline ('_'), and ending with a 2-character upper case ISO 3166-1 alpha 2 country code. (eg. ``-en_US'', ``-en_GB'', ``-de_DE'' or ``-fr_FR''. Note that ``GB'', and not ``UK'' is the code for Great Britain, although ``UK'' should be recognized for compatibility reasons.) The suffix is included when sorting the tags alphabetically, so the default locale (with no tag-name suffix) always comes first. If the country is unknown or not applicable, a country code of ``XX'' should be used.
Tags with numerical values may allow units of measurement to be specified. The units string is stored in brackets at the end of the tag name, and is composed of zero or more ASCII characters in the range 0x21 to 0x7d, excluding the bracket characters 0x28 and 0x29. (eg. ``Resolution(/cm)'' or ``SpecificHeat(J/kg.K)''.) See Image::ExifTool::MIEUnits for details. Unit strings are not localized, and may not be used in combination with localized text strings.
Sets of tags which would require a common prefix should be added in a separate MIE group instead of adding the prefix to all tag names. For example, instead of these TagName's:
ExternalFlashType ExternalFlashSerialNumber ExternalFlashFired
one would instead designate a separate ``ExternalFlash'' MIE group to contain the following elements:
Type SerialNumber Fired
These extended DataLength fields exist only if DataLength is 255, 254 or 253, and are respectively 2, 4 or 8 byte unsigned integers giving the data block length. One of these values must be used if the data block is larger than 252 bytes, but they may be used if desired for smaller blocks too (although this may add a few unnecessary bytes to the MIE element).
The data value for the MIE element. The format of the data is given by the FormatCode. For MIE group elements, the data includes all contained elements and the group terminator.
A MIE group element is identified by a format code of 0x10 (big endian byte ordering) or 0x18 (little endian). The group terminator is distinguished by a zero TagLength (it is the only element allowed to have a zero TagLength), and has a FormatCode of 0x00.
The MIE group element is permitted to have a zero DataLength only if the data is uncompressed. This special value indicates that the group length is unknown (otherwise the minimum value for DataLength is 4, corresponding the the minimum group size which includes a terminator of at least 4 bytes). If DataLength is zero, all elements in the group must be parsed until the group terminator is found. If non-zero, DataLength includes the length of all elements contained within the group, including the group terminator. Use of a non-zero DataLength is encouraged because it allows readers quickly skip over entire MIE groups. For compressed groups DataLength must be non-zero, and is the length of the compressed group data (which includes the compressed group terminator).
The group terminator has a FormatCode and TagLength of zero. The terminator DataLength must be 0, 6 or 10 bytes, and extended DataLength codes may not be used. With a zero DataLength, the byte sequence for a terminator is ``7e 00 00 00'' (hex). With a DataLength of 6 or 10 bytes, the terminator data block contains information about the length and byte ordering of the preceding group. This additional information is recommended for file-level groups, and is used in multi-document MIE files and MIE trailers to allow the file to be scanned backwards from the end. (This may also allow some documents to be recovered if part of the file is corrupted.) The structure of this optional terminator data block is as follows:
4 or 8 bytes GroupLength (unsigned integer) 1 byte ByteOrder (0x10 or 0x18, same as MIE group) 1 byte GroupLengthSize (0x04 or 0x08)
The ByteOrder and GroupLengthSize values give the byte ordering and size of the GroupLength integer. The GroupLength value is the total length of the entire MIE group ending with this terminator, including the opening MIE group element and the terminator itself.
File-level MIE groups
File-level MIE groups may NOT be compressed.
All elements in a MIE file are contained within a special group with a TagName of ``0MIE''. The purpose of the ``OMIE'' group is to provide a unique signature at the start of the file, and to encapsulate information allowing files to be easily combined. The ``0MIE'' group must be terminated like any other group, but it is recommended that the terminator of a file-level group include the optional data block (defined above) to provide information about the group length and byte order.
It is valid to have more than one ``0MIE'' group at the file level, allowing multiple documents in a single MIE file. Furthermore, the MIE structure enables multi-document files to be generated by simply concatenating two or more MIE files.
This algorithm may be repeated again beginning at this point in the file to locate the next-to-last document, etc.
The table below lists all 5 valid patterns for the last 14 bytes of a file-level MIE group, with all numbers in hex. The comments indicate the length and byte ordering of GroupLength (xx) if available:
?? ?? ?? ?? ?? ?? ?? ?? ?? ?? 7e 00 00 00 - (no GroupLength) ?? ?? ?? ?? 7e 00 00 06 xx xx xx xx 10 04 - 4 bytes, big endian ?? ?? ?? ?? 7e 00 00 06 xx xx xx xx 18 04 - 4 bytes, little endian 7e 00 00 0a xx xx xx xx xx xx xx xx 10 08 - 8 bytes, big endian 7e 00 00 0a xx xx xx xx xx xx xx xx 18 08 - 8 bytes, little endian
?? ?? ?? ?? 7e 00 04 00 7a 6d 69 65 7e 00 00 06 xx xx xx xx 10 04 ?? ?? ?? ?? 7e 00 04 00 7a 6d 69 65 7e 00 00 06 xx xx xx xx 18 04 7e 00 04 00 7a 6d 69 65 7e 00 00 0a xx xx xx xx xx xx xx xx 10 08 7e 00 04 00 7a 6d 69 65 7e 00 00 0a xx xx xx xx xx xx xx xx 18 08
Note that the zero-DataLength terminator may not be used here because the trailer length must be known for seeking backwards from the end of the file.
Multiple trailers may be appended to the same file using this technique.
It is preferred that closely related values with the same format are written to a single tag instead of using multiple tags. This improves localization of like values and decreases MIE element overhead. For instance, instead of separate ImageWidth and ImageHeight tags, a single ImageSize tag is defined.
Tags which may take on a discrete set of values should have meaningful values if possible. This improves the extensibility of the format and allows a more reasonable interpretation of unrecognized values.
Integer and floating point numbers may be represented in binary or string form. In string form, integers are a series of digits with an optional leading sign (eg. ``[+|-]DDDDDD''), and multiple values are separated by a single space character (eg. ``23 128 -32''). Floating point numbers are similar but may also contain a decimal point and/or a signed exponent with a leading 'e' character (eg. ``[+|-]DD[.DDDDDD][e(+|-)EEE]''). The string ``inf'' is used to represent infinity. One advantage of numerical strings is that they can have an arbitrarily high precision because the possible number of significant digits is virtually unlimited.
Note that numerical values may have associated units of measurement which are specified in the ``TagName'' string.
All MIE dates are strings in the form ``YYYY:mm:dd HH:MM:SS.ss+HH:MM''. The fractional seconds (``.ss'') are optional, and if included may contain any number of significant digits (unlike all other fields which are a fixed number of digits and must be padded with leading zeros if necessary). The timezone (``+HH:MM'' or ``-HH:MM'') is recommended but not required. If not given, the local system timezone is assumed.
exiftool -o new.mie -tagsfromfile FILE '-mie:all<all' \ '-subfilename<filename' '-subfiletype<filetype' \ '-subfilemimetype<mimetype' '-subfiledata<=FILE'
where "FILE" is the name of the file.
For unrecognized files, this command may be used:
exiftool -o new.mie -subfilename=FILE -subfiletype=TYPE \ -subfilemimetype=MIME '-subfiledata<=FILE'
where "TYPE" and "MIME" represent the source file type and MIME type respectively.
Creating a MIE trailer with ExifTool is a two-step process since ExifTool can't currently be used to add a MIE trailer directly. The example below illustrates the steps for adding a MIE trailer with a small preview image ("small.jpg") to a destination JPEG image ("dst.jpg").
Step 1) Create a MIE file with a TrailerSignature containing the desired information:
exiftool -o new.mie -trailersignature=1 -tagsfromfile small.jpg \ '-previewimagetype<filetype' '-previewimagesize<imagesize' \ '-previewimagename<filename' '-previewimage<=small.jpg'
Step 2) Append the MIE information to another file. In Unix, this can be done with the 'cat' command:
cat new.mie >> dst.jpg
Once added, ExifTool may be used to edit or delete a MIE trailer in a JPEG or TIFF image.
# write the Author tag in the second MIE document exiftool -mie2:author=phil test.mie # delete the first MIE document from a file exiftool -mie1:all= test.mie
exiftool -mie:gpsaltitude='7500(ft)' test.mie
If no units are provided, the default units are written.
exiftool -comment-en_us='this is a comment' test.mie
2010-04-05 - Fixed "Format Size" Note 7 to give the correct number of bits in the example rational value 2007-01-21 - Specified LF character (0x0a) for text newline sequence 2007-01-19 - Specified ISO 8859-1 character set for extended ASCII codes 2007-01-01 - Improved wording of Step 5 for scanning backwards in MIE file 2006-12-30 - Added EXAMPLES section and note about UTF BOM 2006-12-20 - MIE 1.1: Changed meaning of TypeModifier bit (0x08) for unknown data (FormatType 0x00), and documented byte swapping 2006-12-14 - MIE 1.0: Added Data Values and Numerical Representations sections, and added ability to specify units in tag names 2006-11-09 - Added Levels of Support section 2006-11-03 - Added Trailer Signature 2005-11-18 - Original specification created
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself. The MIE format itself is also copyright Phil Harvey, and is covered by the same free-use license.