Skip to content

Latest commit

 

History

History
110 lines (96 loc) · 5.31 KB

WHY.md

File metadata and controls

110 lines (96 loc) · 5.31 KB

Why Should I use Elektra?

The three main points relevant for most people are:

  1. Even though Elektra provides a global key database, that allows read- and write access of every single parameter for any application in an integrated fashion, configuration files stay human read- and writable.
  2. Flexible adoption on how the configuration settings are accessed via plugins: you can run arbitrary code in multiple languages or notify others when configuration files are changed. Plugins allow us to support hundreds of different configuration file formats.
  3. Elektra allows us to specify configuration values:

Why not Other Solutions?

Some might ask: isn't this solution overkill? Why not tackle these three issues separately? The answer is:

  1. If we would only implement a configuration file library for applications, we would hinder the work of administrators and would not provide external access to configuration settings or specifications.
  2. If we only implement an administrator tool that can parse and generate configuration files, but is not used by the applications themselves, we create a gap that leads to inconsistent understanding of the configuration file syntax.
  3. If we only specify the meaning of configuration settings within applications but not in a form accessible for administrators, we would create a gap that leads to inconsistent understanding of the configuration settings' semantics.

For common understanding of syntax and semantics of configuration files a full-stack solution like Elektra is required. We acknowledge, however, that such a change cannot be done overnight, thus we integrate as many configuration file formats as possible. This way, people can continue using files in /etc, regardless of whether or not Elektra is used.

To give one example, in OpenLDAP 2.4.39 the value of listener-threads will be reduced to the next number that is a power of 2. To correctly manipulate the setting we need not only know the syntax of how to write listener-threads correctly in the configuration file, but also the knowledge how the value is transformed internally. Elektra solves all three issues, and then users can easily externally manipulate listener-threads, without caring about the concrete syntax of the file and getting feedback of the semantics (you might get validation errors and you can receive the value exactly as the application will get it).

Unique Features

Features that rarely can be found elsewhere (at least in this combination):

  • Bootstrap code and proper abstraction is included:
    • You do not need to worry about the file names of configuration files in the application.
    • Cascading between /etc, $HOME, cwd and so on.
    • You can change which Elektra path is connected to which configuration file with mounting.
    • Portable across OS (Linux, BSD, w64, macOS,.. ) and desktop systems (GNOME, KDE,...).
  • No daemon, so no single point of failure but still having guarantees of consistent, validated files with good performance.
  • Provides 3-way merging for configuration upgrades.

Further Reasons

  • Links and automatic calculation of values: unlike with other solutions you do not need to duplicate configuration values for different applications but you can comfortably link between them which makes many inconsistencies impossible.
  • Allows us to easily create GUIs and web-UIs for the whole configuration settings on the system.
  • Allows you to import/export all parts of the configuration settings.
  • Syntax independence: you can consistently use your favorite syntax.
  • Configuration Management (such as Puppet) can be used on top of it without having to fiddle with specifics of every configuration file.
  • CLI-tool available
  • kdb editor allows you to edit any path of Elektra with your favorite syntax (independent of the actual syntax of the configuration files that store values of this path).
  • Allows us to also integrate command-line arguments and environment into a consistent place for configuration.
  • Reduces huge amount of code: Nearly every application has very similar code:
    • finding the correct configuration file (for different OS)
    • parsing configuration files
    • validating configuration files
    • replace configuration files on changes
  • All advantages libraries have:
    • Performance: Improvements in the library benefits all applications.
    • The library only needs to be loaded once in the memory.
    • On fixes not all binaries of all applications need to be replaced.
  • All advantages maintained code with a community has:
    • If something does not work, open an issue.
    • If you have a question, open an issue.
    • Regular releases.

Further Readings