Here, my summary of key features and decisions of Guix:
- Guix is a package manager that can (optionally) run on top of Linux distributions or other POSIX systems, like cargo, pip, conda or Conan. In difference to the pip and cargo package managers, it is language-agnostic, supports many different build systems and languages, and features around 29000 packages now.
- Guix allows to define a fully reproducible system. This works by using a declarative language for immutable version-controlled package descriptions, and by deriving any software from package definitions and a fixed version (commit hash) of the source code. In that, it is similar but much stricter than Nix and NixOS. The key point is that any software built, and all its dependencies, go back to unambigously, immutable versions of source code and build recipes - and all inputs to the system are open source and can be reviewed.
- Important for programming, this can also define isolated build and development environments, like Python’s venv, but also Docker containers. This means that Guix can be used to develop, build, package, and deploy software, very much like Snap packages. And that’s independent from the distribution you work in, very much like pip or cargo are independent from the system you work in. (And yes, it supports Rust!).
- This allows it, and also makes it technically possible, that any software package can be re-built and run years later. To make this legally possible, the official distribution of Guix also demands all components to be open source (FOSS). This is also a key difference to NixOS and non-free forks of Guix, which allow non-free binary packages, but sacrifice reproducibility. (To illustrate: If you have a binary, proprietary scanner driver in NixOS, and the owning company practices planned obselescence and decides that you should buy their new hardware, and pulls that driver, you are out of luck. In Guix, this can’t happen.) (Note that as your own private conponents, you can define any package you like, you can also distribute your definitions as a complement to GNU Guix. Non-free packages for Guix do exist, in the same way as you can buy and run Steam Games software for Linux. Such non-free software just can’t become part of the official Guix distribution, just like Amazon or Apple can’t sell their non-free software via Debian or the Linux kernel project (or, for that matter, Apple has no obligation to market and distribute, say, Oracle products).
- All inputs being open source also means that any software component can be reviewed, that mis-features such as privacy-invasive behaviour can be removed, and that it is hardly possible to hide malware in the system. Because this also applies recursively to all compilers and build tools, this solves also Thompson’s “Trusting Trust” problem. In fact, the whole system can be build from a 512 byte binary root (called MER). (Interestingly, that level of user control gets a lot of hate online – certain companies don’t seem to like it).
- Because it would take too long to build every user package from source every time, the produced packages are normally cached (while their correct binary content can be easily verified).
- The declarative description language for the packages is a well-defined, established, minimalist language called Scheme. This is a member of the Lisp family of languages. That Lisp is very well suited for declaratively building and configuring large systems has been proven with GNU Emacs, whose software, but more importantly, whole user configuration, is written in Emacs Lisp.
- The Scheme implementation used is called Guile. It has especially good support for the POSIX environment and has also much better-than-average interactive debugging capabilities compared to other Scheme implementations.
- Also worth noting is that the Guix project has superb online documentation. This is a practical advantage compared to Nix.
As example: you are on Debian stable and quickly want to try a recent version of the kakoune editor (as kakoune is in ongoing development): They are available under the Guix package manager. Just
guix install kakoune
and bang you have it!
How it works:
https://codeberg.org/guix/guix#headline-4
Manual:
https://guix.gnu.org/manual/en/html_node/Installation.html
Also informative for using Guix just as a package manager:
This comment perfect captures why I don’t like Lisp. Essentially “it’s simple, this easy to read code transforms to this AST”. Lisp basically says “we can make parsing way easier if we force programmers to write the AST directly!” which is really stupid because computers can perfectly well parse syntax that is easy for humans to read and turn it into ASTs automatically.
It makes it easier to parse for computers at the cost of being much harder to parse for humans, which is really the wrong choice in most cases. (The exception is if you’re DIYing your compiler, e.g. if you’re teaching how to write a compiler then Lisp is a good target.)
I program in a lot of languages, including C, C++, Rust, Python, Java, and Lisps are not more difficult than the others. In fact, they tend to be more orthogonal: Scheme can express mote things than Python, e.g. with its continuations, but is much simpler and minimal at the same time. In Python, you have list comprehensions, generator expressions, generators, iterators, dictionary comprehensions - in Clojure you have lists and lazy sequences, period. Guile and Common Lisps like SBCL compile to native code, and are a lot faster than Python - SBCL is about as fast as modern Java - and so on.
In addition, what your comment distorts is that C, for example, was conceived as “portable assembly”, while Lisps S-expressions were originally conceived as mathematical notation for algorithms - it was only later discovered that they can be compiled down to native code.
But as I already mentioned, when it comes to Guix, one sees a lot of hate, misconceptions, and trolling, and one can ask why.
Right, I’m not saying it isn’t simpler in terms of syntax. The point I was making is that the syntax is simpler but in a way that makes it worse because while it’s easier for computers to read, it’s harder for humans.
That sounds extremely unlikely. I think you’re misinterpreting this quote (which is fair enough; it’s not very clear):
As far as I can tell Lisp was always intended to be compiled and executed. That quote is about compiling the
eval()function (which was just meant to explain how Lisp is executed) into a binary and using that as an interpreter.Also I skimmed the paper that is from, and in fact Lisp was intended to be targeted by AI (in the same way that we get AI to write and execute Python to solve problems), which explains a lot. It wasn’t designed for humans to write, so why bother with nice syntax; just have the machine write the AST directly!
(I expect that was only part of the motivation tbf, but still!)