The comparison table is ... inacurate, to say the least. I mostly know about Nix, and it is to large extend toolchain (gcc vs. llvm) and libc (glibc vs. musl vs. baremetal and niche stuff) agnostic. And claiming it has a centralized trust model is, well, also a stretch. I'd rather claim it has no thorough trust model but is very friendly towards detailed introspection of everything.
Maybe, the comparison table should say nixpkgs, instead of nix, if that's what you mean?
To the best of my knowledge, most of nixpkgs is built with gcc/glibc.
Not totally sure on the trust model, but from the options they give, centralised (nixpkgs) or decentralised (all maintainers of nixpkgs) sounds about right... I would see how nixpkgs is one system
That is the CLI, in Nix expressions itself it is similarly simple to choose. Of course you can also build for let's say android, than you might end up with a bionic libc. Point in case: nixpkgs really abstracts over this.
Not totally sure on the trust model, but from the options they give, centralised (nixpkgs) or decentralised (all maintainers of nixpkgs) sounds about right... I would see how nixpkgs is one system
But it is developed by hundres if not thousands of contributors, and you can always opt out of the binary cache and build from source yourself.
Actually this yields a GCC / musl version of hello. pkgsMusl.pkgsLLVM.hello yields clang / musl, and for a good measure the clang used is also linked against musl. You can control what is going on, it's just that the way these overlays work has some catches in the example you have given.
I kind of hoped that they are trying to maintain a double stack to minimise the number of upstreams that are irreplaceableā¦ but no, they seem to be just about code provenance tracking and bootstrap inspectability.
wucke13 | a day ago
The comparison table is ... inacurate, to say the least. I mostly know about Nix, and it is to large extend toolchain (gcc vs. llvm) and libc (glibc vs. musl vs. baremetal and niche stuff) agnostic. And claiming it has a centralized trust model is, well, also a stretch. I'd rather claim it has no thorough trust model but is very friendly towards detailed introspection of everything.
fosslinux | a day ago
Maybe, the comparison table should say nixpkgs, instead of nix, if that's what you mean? To the best of my knowledge, most of nixpkgs is built with gcc/glibc.
Not totally sure on the trust model, but from the options they give, centralised (nixpkgs) or decentralised (all maintainers of nixpkgs) sounds about right... I would see how nixpkgs is one system
k749gtnc9l3w | a day ago
The main Linux target for Nixpkgs is
gccandglibc, butpkgsMuslandpkgsLLVMin-repo overlays exist and have some package coverage.wucke13 | 21 hours ago
... per default. However, that is easily overriden:
That is the CLI, in Nix expressions itself it is similarly simple to choose. Of course you can also build for let's say android, than you might end up with a bionic libc. Point in case: nixpkgs really abstracts over this.
But it is developed by hundres if not thousands of contributors, and you can always opt out of the binary cache and build from source yourself.
k749gtnc9l3w | 19 hours ago
Actually this yields a GCC / musl version of
hello.pkgsMusl.pkgsLLVM.helloyields clang / musl, and for a good measure the clang used is also linked against musl. You can control what is going on, it's just that the way these overlays work has some catches in the example you have given.wucke13 | 14 hours ago
Good catch, I vaguely remember having ran into this some time ago!
kghose | a day ago
What is this for? Is it for images for mass deployments?
David_Gerard | a day ago
looking at the sponsors, it's for cryptocurrency. ("Decentralized" is a bit of a red flag which led me to look at who the sponsors were.)
gcupc | a day ago
The website is giving LLM... looks like a lot of websites I've seen the last few weeks.
sknebel | a day ago
I don't think it strictly has a specific target like that, although servers/appliances in the widest sense seem maybe more likely than desktops?
pyfisch | a day ago
I was surprised to see that Stagex uses Codeberg as its forge: https://codeberg.org/stagex/stagex
k749gtnc9l3w | a day ago
I kind of hoped that they are trying to maintain a double stack to minimise the number of upstreams that are irreplaceableā¦ but no, they seem to be just about code provenance tracking and bootstrap inspectability.