 | ||
A filename (also written as two words, file name) is a name used to uniquely identify a computer file stored in a file system. Different file systems impose different restrictions on filename lengths and the allowed characters within filenames.
Contents
- History
- Unicode migration
- Number of names per file
- Length restrictions
- Filename extensions
- Encoding interoperability
- Encoding indication interoperability
- Unicode interoperability
- Perspectives
- Uniqueness
- Letter case preservation
- Reserved characters and words
- References
A filename may include one or more of these components:
The components required to identify a file varies across operating systems, as does the syntax and format for a valid filename.
Discussions of filenames are complicated by a lack of standardisation of the term. Sometimes "filename" is used to mean the entire name, such as the Windows name c:directorymyfile.txt. Sometimes, it will be used to refer to the components, so the filename in this case would be myfile.txt. Sometimes, it is a reference that excludes an extension, so the filename would be just myfile. Such ambiguity is widespread and this article does not attempt to define any one meaning, and indeed may be using any of these meanings. Some systems will adopt their own standardised nomenclature like "path name", but these too are not standardised across systems.
History
Around 1962, the Compatible Time-Sharing System introduced the concept of a file (i.e. non-paper file).
Around this same time appeared the dot (period or full-stop) as a filename extension separator, and the limit to three letter extensions might have come from RAD50 16-bit limits.
Traditionally, filenames allowed only alphanumeric characters, but as time progressed, the number of characters allowed increased. This led to compatibility problems when moving files from one file system to another.
Around 1995, VFAT, an extension to the FAT filesystem, was introduced in Windows 95 and Windows NT 3.5. It allowed mixed-case Unicode long filenames (LFNs), in addition to classic "8.3" names.
In 1985, RFC 959 officially defined a pathname to be the character string that must be entered into a file system by a user in order to identify a file.
Unicode migration
One issue was migration to Unicode. For this purpose, several software companies provided software for migrating filenames to the new Unicode encoding.
Mac OS X 10.3 marked Apple's adoption of Unicode 3.2 character decomposition, superseding the Unicode 2.1 decomposition used previously. This change caused problems for developers writing software for Mac OS X.
Number of names per file
Unix-like file systems allow a file to have more than one name; in traditional Unix-style file systems, the names are hard links to the file's inode or equivalent. Windows supports hard links on NTFS file systems, and provides the command fsutil in Windows XP, and mklink in later versions, for creating them. Hard links are different from Windows shortcuts, classic Mac OS/macOS aliases, or symbolic links. The introduction of LFNs with VFAT allowed filename aliases. For example, longfi~1.??? with a maximum of eight plus three characters was a filename alias of "long file name.???" as a way to conform to 8.3 limitations for older programs.
This property was used by the move command algorithm that first creates a second filename and then only removes the first filename.
Other filesystems, by design, provide only one filename per file, which guarantees that alteration of one filename's file does not alter the other filename's file.
Length restrictions
Some filesystems restrict the length of filenames. In some cases, these lengths apply to the entire file name, as in 44 characters on IBM S/370. In other cases, the length limits may apply to particular portions of the filename, such as the name of a file in a directory, or a directory name. For example, 9 (e.g., 8-bit FAT in Standalone Disk BASIC), 11 (e.g. FAT12, FAT16, FAT32 in DOS), 14 (e.g. early Unix), 21 (Human68K), 31, 30 (e.g. Apple DOS 3.2 and 3.3), 15 (e.g. Apple ProDOS), 44 (e.g. IBM S/370), or 255 (e.g. early Berkeley Unix) characters or bytes. Length limits often result from assigning fixed space in a filesystem to storing components of names, so increasing limits often requires an incompatible change, as well as reserving more space.
A particular issue with filesystems that store information in nested directories is that it may be possible to create a file whose total name exceeds implementation limits, since length checking may apply only to individual parts of the name rather than the entire name. Many Windows applications are limited to a MAX_PATH value of 260, but Windows file names can easily exceed this limit [1].
Filename extensions
Many file systems, including FAT, NTFS, and VMS systems, allow a filename extension that consists of one or more characters following the last period in the filename, dividing the filename into two parts: a base name or stem and an extension or suffix used by some applications to indicate the file type. Multiple output files created by an application use the same basename and various extensions. For example, a compiler might use the extension FOR for source input file (for Fortran code), OBJ for the object output and LST for the listing. Although there are some common extensions, they are arbitrary and a different application might use REL and RPT. On filesystems that do not segregate the extension, files will often have a longer extension such as html.
Encoding interoperability
There is no general encoding standard for filenames.
Because file names have to be exchanged between software environments (think network file transfer, file system storage, backup and file synchronization software, configuration management, data compression and archiving, etc.), it is very important not to lose file name information between applications. This led to wide adoption of Unicode as a standard for encoding file names, although legacy software might be non-Unicode-aware.
Encoding indication interoperability
Traditionally, filenames allowed any character in their filenames as long as they were file system safe. Although this permitted the use of any encoding, and thus allowed the representation of any local text on any local system, it caused many interoperability issues.
A filename could be stored using different byte strings in distinct systems within a single country, such as if one used Japanese Shift JIS encoding and another Japanese EUC encoding. Conversion was not possible as most systems did not expose a description of the encoding used for a filename as part of the extended file information. This forced costly filename encoding guessing with each file access.
A solution was to adopt Unicode as the encoding for filenames.
In the classic Mac OS, however, encoding of the filename was stored with the filename attributes.
Unicode interoperability
The Unicode standard solves the encoding determination issue.
Nonetheless, some limited interoperability issues remain, such as normalization (equivalence), or the Unicode version in use. For instance, UDF is limited to Unicode 2.0; macOS applies NFD Unicode normalization and is optionally case-sensitive (case-insensitive by default.) Filename maximum length is not standard and might depend on the code unit size. Although it is a serious issue, in most cases this is a limited one.
On Linux, this means the filename is not enough to open a file: additionally, the exact byte representation of the filename on the storage device is needed. This can be solved at the application level, with some tricky normalization calls.
The issue of Unicode equivalence is known as "normalized-name collision". A solution is the Non-normalizing Unicode Composition Awareness used in the Subversion and Apache technical communities. This solution does not normalize paths in the repository. Paths are only normalized for the purpose of comparisons. Nonetheless, some communities have patented this strategy, forbidding its use by other communities.
Perspectives
To limit interoperability issues, some ideas described by Sun are to:
Those considerations create a limitation not allowing a switch to a future encoding different from UTF-8.
Uniqueness
Within a single directory, filenames must be unique. Since the filename syntax also applies for directories, it is not possible to create a file and directory entries with the same name in a single directory. Multiple files in different directories may have the same name.
Uniqueness approach may differ both on the case sensitivity and on the Unicode normalization form such as NFC, NFD. This means two separate files might be created with the same text filename and a different byte implementation of the filename, such as L"x00C0.txt" (UTF-16, NFC) (Latin capital A with grave) and L"x0041x0300.txt" (UTF-16, NFD) (Latin capital A, grave combining).
Letter case preservation
Some filesystems, such as FAT, store filenames as upper-case regardless of the letter case used to create them. For example, a file created with the name "MyName.Txt" or "myname.txt" would be stored with the filename "MYNAME.TXT". Any variation of upper and lower case can be used to refer to the same file. These kinds of file systems are called case-insensitive and are not case-preserving. Some filesystems prohibit the use of lower case letters in filenames altogether.
Some file systems store filenames in the form that they were originally created; these are referred to as case-retentive or case-preserving. Such a file system can be case-sensitive or case-insensitive. If case-sensitive, then "MyName.Txt" and "myname.txt" may refer to two different files in the same directory, and each file must be referenced by the exact capitalisation by which it is named. On a case-insensitive, case-preserving file system, on the other hand, only one of "MyName.Txt", "myname.txt" and "Myname.TXT" can be the name of a file in a given directory at a given time, and a file with one of these names can be referenced by any capitalisation of the name.
From its original inception, Unix and its derivative systems were case-preserving. However, not all Unix-like file systems are case-sensitive; by default, HFS+ in macOS is case-insensitive, and SMB servers usually provide case-insensitive behavior (even when the underlying file system is case-sensitive, e.g. Samba on most Unix-like systems), and SMB client file systems provide case-insensitive behavior. File system case sensitivity is a considerable challenge for software such as Samba and Wine, which must interoperate efficiently with both systems that treat uppercase and lowercase files as different and with systems that treat them the same.
Reserved characters and words
File systems have not always provided the same character set for composing a filename. Before Unicode became a de facto standard, file systems mostly used a locale-dependent character set. By contrast, some new systems permit a filename to be composed of almost any character of the Unicode repertoire, and even certain non-Unicode byte sequences. Limitations may be imposed by the file system, operating system, application, or requirements for interoperability with other systems.
Many file system utilities prohibit control characters from appearing in filenames. In Unix-like file systems, the null character and the path separator / are prohibited.
File system utilities and naming conventions on Windows prohibit particular characters from appearing in filenames:
Note 1: While they are allowed in Unix file and folder names, most Unix shells require certain characters such as spaces, <, >, |, , and sometimes :, (, ), &, ;, #, as well as wildcards such as ? and *, to be quoted or escaped:
five and six<seven (example of escaping)'five and six<seven' or "five and six<seven" (examples of quoting)
The character 0xE5 was not allowed as the first letter in a filename under 86-DOS and MS-DOS/PC DOS 1.x-2.x, but can be used in later versions.
In Windows utilities, the space and the period are not allowed as the final character of a filename. The period is allowed as the first character, but certain Windows applications, such as Windows Explorer, forbid creating or renaming such files (despite this convention being used in Unix-like systems to describe hidden files and directories). Workarounds include appending a dot when renaming the file (that is then automatically removed afterwards), using alternative file managers, or saving a file with the desired filename from within an application.
Some file systems on a given operating system (especially file systems originally implemented on other operating systems), and particular applications on that operating system, may apply further restrictions and interpretations. See comparison of file systems for more details on restrictions.
In Unix-like systems, DOS, and Windows, the filenames "." and ".." have special meanings (current and parent directory respectively). Windows 95/98/ME also uses names like "...", "...." and so on to denote grandparent or great-grandparent directories. All Windows versions forbid creation of filenames that consist of only dots, although names consist of three dots ("...") or more are legal in Unix.
In addition, in Windows and DOS utilities, some words are also reserved and cannot be used as filenames. For example, DOS device files:
CON, PRN, AUX, CLOCK$, NUL COM1, COM2, COM3, COM4 LPT1, LPT2, LPT3, LPT4 (LPT4 only in some versions of DR-DOS) LST (only in 86-DOS and DOS 1.xx) KEYBD$, SCREEN$ (only in multitasking MS-DOS 4.0) $IDLE$ (only in Concurrent DOS 386, Multiuser DOS and DR DOS 5.0 and higher) CONFIG$ (only in MS-DOS 7.0-8.0)Systems that have these restrictions cause incompatibilities with some other filesystems. For example, Windows will fail to handle, or raise error reports for, these legal UNIX filenames: aux.c, q"uote"s.txt, or NUL.txt.
NTFS filenames that are used internally include:
$Mft, $MftMirr, $LogFile, $Volume, $AttrDef, $Bitmap, $Boot, $BadClus, $Secure, $Upcase, $Extend, $Quota, $ObjId and $Reparse