FAT Filenames

Format

DOS uses the following pattern for forming file paths:

[disk]:\[directory]\[subdirectories]\[filename].[extension]

[disk] is one letter of the Latin alphabet. Valid letters are A through Z. How disk letters are assigned to physical devices was discussed earlier.

[directory] and [subdirectories] are strings. They specify the location of the file in the directory tree. None, either or both are omitted depending on file location.

[filename] is also a string that identifies name of the file. [extension] usually contains some information regarding type of the file. It is a string. Note that directories may also have extensions. Because file and directory names are of the same format, everything below that applies to filenames also applies to directory names.

Short and Long Names

• Letters A through Z.
• Digits 0 through 9.
• Characters with ASCII codes greater than 127. Note that these characters depend on current code table. Also, their handling is case sensitive.
• Space. Note that many applications do not recognize this character and it is not used in creating aliases for long filenames.
• $ % - _ @ ~ ` ! ( ) { } ^ # &

• . (dot) is the delimiter between [name] and [extension]. There should be only one delimiter for each file or directory name.
• \ (back slash) is the delimiter between directory and file names.
• : (colon) is the delimiter between disk letter and the rest of the path.

• ? (question mark) denotes any character.
• * (asterisk) denotes blank or any number of any characters.

Long names are up to 256 characters long, including extension and the terminating null character. This limitation is artificial, and the long names on the disk can actually be longer than 256 characters. Again, some limitations were created by the software that serves FAT. The maximum length of a directory path, including drive letter, column, and leading slash, but excluding trailing slash, null terminator, filename and extension, is 246 bytes. The maximum length of a full path is 260 characters, including null terminator. This is four characters more than I recommend.

Long filenames are stored in unicode. Each character is two bytes long. There are two important things to remember about unicode:

• For only the characters in the low half of the ASCII table, the high byte of unicode character is zero, and the low byte is the same as in ASCII.
• Unicode system supports case insensitive operations for all characters, not just Latin.

• + , ; = [ ]
• . (dot) can occur more than once in a filename. Extension is the substring after the last dot.

Aliasing

• If one or more of the characters are not upper case, the alias created is the same as the long filename, only converted to upper case.
• If all characters are uppercase, then only the alias is created, but the directory entry for the long filename itself is not created.

• First, illegal for short filenames characters are deleted from long filename. All dots except the last one are also deleted.
• Then, first six (or less, if the name is shorter) characters are used as base, and ~ (tilde) and the number 1 are used as tail. Base and tail are concatenated to form the alias' filename. Alias' extension is formed by the first three (or less) letters from the long filename's extension.
• If the alias with such a name already exists, the number in the tail is increased. This process is repeated until we arrive at a unique alias. Note that when the number reaches the power of ten, its decimal representation takes up one more byte. In this case one last character is stripped off the base. The situation when this loop overflows (base is empty, but number needs to be increased) is very unlikely. After all, where did you see more than 9,999,999 files in one directory?

One can access the file using its alias.

• When the file is created using its alias, the corresponding long name is not created, with the exception of one case below.
• When the file is deleted, either via its alias or via its long name, all directory entries for the long name and the alias are marked as deleted.
• Some extra steps are taken to preserve the long filename. The long filename is obviously lost when the file is copied using its alias. But there are less obvious things. For example, an application may delete and re-create the file using its alias. Not to loose the long filename, Windows 95 keeps the information about the deleted via aliases files for fifteen seconds, and if an an attempt was made to re-create the file during this time using the same alias, the old long filename is associated with this alias.