Functional package management (FPM) is the next step forward [[https://edolstra.github.io/pubs/nixos-icfp2008-final.pdf][pioneered by Eelco Dolstra]] with [[id:3b3fdcbf-eb40-4c89-81f3-9d937a0be53c][Nix package manager]] as the living implementation. [fn:: Eventually, it became the inspiration for other projects that focuses on reproducibility and security such as [[https://guix.gnu.org/][GNU Guix]] and [[https://silverblue.fedoraproject.org/][Fedora Silverblue]].]
The conceptual overview of it is similar to [[id:7e1e2bcc-6f70-4624-a049-b42f9db5e28b][Functional programming]] with how it views with data and functions (in this case, packages and build processes).
It views that packages are a unique result from combining different things: dependencies, build processes, versions, and more.
If a dependency has updated its patch version or a build instruction has revised with one line of change, it will result in a "new" package as the output even if the package isn't any different.
Coming from a functional approach, it also features statelessness and immutability, making it a reliable toolkit for [[https://reproducible-builds.org/][reproducible builds]] (thus gaining a reach for [[id:6eeb7a24-b662-46d6-9ece-00a5028ff4d8][Reproducible research]]).
- Multiple versions of packages are hard to place side-by-side which is used by developers targetting stable versions while experimenting with the latest versions.
- Destructive ongoing updates which consists of replacing the current version with the updated version while the update is going.
The functional approach with the unique output and immutable variables addresses the listed problems in some way.
- A package is built from an input resulting in a package on a unique immutable location, eliminating the statefulness in the process.
- Since each revision results in a new location, the same package with different versions can be easily placed side-by-side.
- With the way how packages built, the old system can be kept giving the possibility for the user to rollback to those previous versions.
- The functional approach creates a guaranteed output, making it predictable (and traceable).