Some applications provide tools to create, modify, and verify the syntax of their configuration files; these sometimes have graphical interfaces. For other programs, system administrators may be expected to create and modify files by hand using a text editor, which is possible because many are human-editable plain text files. For server processes and operating-system settings, there is often no standard tool, but operating systems may provide their own graphical interfaces such as YaST or debconf.
Some computer programs only read their configuration files at startup. Others periodically check the configuration files for changes. Users can instruct some programs to re-read the configuration files and apply the changes to the current process, or indeed to read arbitrary files as a configuration file. There are no definitive standards or strong conventions.
A configuration file for GNU GRUB being edited. Comments (the lines beginning with a #) are used both as documentation and as a way to "disable" the setting.
Configuration files and operating systems
Unix and Unix-like operating systems
Across Unix-like operating systems many different configuration-file formats exist, with each application or service potentially having a unique format, but there is a strong tradition of them being in human-editable plain text, and a simple key–value pair format is common. Filename extensions of .cnf, .conf, .cfg, .cf or .ini are often used.
Almost all formats allow comments, in which case, individual settings can be disabled by prepending with the comment character. Often the default configuration files contain extensive internal documentation in the form of comments[1][2] and man files are also typically used to document the format and options available.
System-wide software often uses configuration files stored in /etc, while user applications often use a "dotfile" – a file or directory in the home directory prefixed with a period, which in Unix hides the file or directory from casual listing. Since this causes pollution, newer user applications generally make their own folder in the .config directory, a standardized subdirectory of the home directory.
Some configuration files run a set of commands upon startup. A common convention is for such files to have "rc" in their name,[3] typically using the name of the program then an "(.)rc" suffix e.g. ".xinitrc", ".vimrc", ".bashrc", "xsane.rc". See run commands for further details.
By contrast, IBM's AIX uses an Object Data Manager (ODM) database to store much of its system settings.
MS-DOS
MS-DOS itself primarily relied on just one configuration file, CONFIG.SYS. This was a plain text file with simple key–value pairs (e.g. DEVICEHIGH=C:\DOS\ANSI.SYS) until MS-DOS 6, which introduced an INI-file style format. There was also a standard plain text batch file named AUTOEXEC.BAT that ran a series of commands on boot. Both these files were retained up to Windows 98SE, which still ran on top of MS-DOS.
DOS applications used a wide variety of individual configuration files, most of them binary, proprietary and undocumented - and there were no common conventions or formats.[citation needed]
Microsoft Windows
The REGEDIT application being used to edit Windows Registry data
The early Microsoft Windows family of operating systems heavily utilized plain-text INI files (from "initialization"). These served as the primary mechanism to configure the operating system and application features.[4] The APIs to read and write from these still exist in Windows, but after 1993, Microsoft began to steer developers away from using INI files and toward storing settings in the Windows Registry, a hierarchical database to store configuration settings, which was introduced that year with Windows NT.
IBM's OS/2 uses a binary format, also with a .INI suffix, but this differs from the Windows versions. It contains a list of lists of untyped key–value pairs.[5]
Two files control system-wide settings: OS2.INI and OS2SYS.INI. Application developers can choose whether to use them or to create a specific file for their applications.
HarmonyOS and OpenHarmony operating systems
HarmonyOS and OpenHarmony-based operating systems uses “config.json” configuration file in the root directory of each HAP application. It contains three modules such as app, deviceConfig and Module. The config.json file adheres to the JSON file format, where each entry comprises an attribute and its corresponding value. The sequence of attribute values is arbitrary, but each attribute must be unique and non-repetitive. The value assigned to each attribute aligns with the fundamental data types of JSON. HarmonyOS’ DevEco Studio provides two methods for editing config.json: the code editing view and the visual editing view. The `app` object holds the universal configuration details for the application, as demonstrated in the config.json configuration file of HarmonyOS's foundational application. DeviceConfig encompasses the configuration details for specific devices, including default, Phone, Tablet, PC, TV, Car, Wearable, liteWearable, and smartVision. The settings under the default label apply to all devices. If there are unique requirements for other devices, the settings under the respective device type label need to be configured. The Module object carries the HAP package configuration details, the essential attributes that each Ability must define (such as package name, class name, type, and capabilities provided by Ability), along with the permissions the application needs to access the system or other protected sections of the application. The MainAbility Indicates the name of HAP package’s entrance ability.[6]
Serialization formats
A number of general-purpose serialization formats exist that can represent complex data structures in an easily stored format, and these are often used as a basis for configuration files, particularly in open-source and platform-neutral software applications and libraries. The specifications describing these formats are routinely made available to the public, thus increasing the availability of parsers and emitters across programming languages.
^Syntax typing refers to the use of syntax to designate data types. In languages that allow syntax typing, type declaration will be syntax-based – e.g. true will be a boolean while "true" will be a string – whereas in languages that do not allow syntax typing it will be semantics-based – e.g. true and "true" will be both recognizable as booleans, while microwave and "microwave" will be both recognizable as strings (this will require the parser to have some prior expectations about the type of a particular field, but this is often the case in configuration files).
^Opinions on whether syntax typing in configuration formats is a good or a bad feature vary among authors. with some considering it a disadvantage (see for example What is wrong with TOML § Syntax typing) and others favoring it.
.png)