It’s quite new but I’ve been cooking up some new bootstrapping features with mise which people may find relevant here: https://mise.jdx.dev/bootstrap.html
It’s for things like dotfiles, apt/brew packages, and LaunchAgents/systemd.
EDIT: I feel a little bad having hijacked this, as someone that hears a lot of opinions about devtools I can definitely say chezmoi is a darling of the community and I highly recommend checking it out.
I am quite intrigued. With the sorry state of security, I am doing everything in VMs and have been trying to settle on the best way to setup a new machine. The process is so clunky that I end up defaulting to bigger instances than I should (more pets than cattle).
Being able to centralize this config is far more attractive than having a separate Ansible or pyinfra process.
Edit: The docs on this are very encouraging! However, it is not clear- what is experimental? Anything notably buggy or churning? Any far off features you hope to implement some day?
ALL mise features are experimental until the design is finalized. I don’t do breaking changes except in exceptional circumstances with a long procedure, so experimental means I have the liberty of making them in any release. It is not at all any marker of quality. That said, in practice I don’t make them often.
I’d argue it’s a good time to play with features since if you have any ideas on how it might better fit your workflow I still have the ability to change the design before it’s largely frozen.
That still feels like more external tooling, plus letting the distro make my packaging decisions. I am already using mise, so this is an almost free boon without more complexity.
I saw it the other day in the release notes and I'm definitely setting this up since I already use mise for many of my global tools and projects.
Some months ago I ported part of my git bare repo to chezmoi but never really picked it up since, even if I see it as a great and very complete tool for managing complex dotfiles, I don't have many machines and can live by with a bunch of if uname -s, and the fact that I still need to wrangle brewfiles and scripts for packages. Having the sytem package managers glue along the configs and symlinks is exactly what I wanted.
Just started moving things to mise and didn’t see this before, thanks.
Hopefully can use this alone instead of needing to combine w/ chezmoi / nix to get everything shell and pkg manager agnostic, consistent, and DRY (bash/zsh/fish + macports/pkgsrc/brew).
it's dependency free. You don't even need brew to install brew formulas.
The "exception" to that are linux package managers like apt-get and dnf which it calls under the hood. I think can't be an actual issue since it's not like you would ever use ubuntu/redhat without their system package manager installed.
Thanks so much for your work on mise! I used to be a heavy asdf user but nowadays I'm an even heavier mise user!
Random question while you're here: mise is undergoing pretty heavy development these days and I recently noticed that 1) my coworkers and I are not always on the same version, so some features/bug fixes are not available to everyone, and 2) package registries often don't have the latest mise version.
So I think we need a meta tool manager here to manage the tool manager version. :) Seriously, though, have you considered having mise manage its own version? I think that'd be pretty neat!
Thinking aloud, I guess one way to do this might be to distribute through package registries only a lightweight bootstrap application, which 1) reads the pinned mise version from mise.toml and downloads it as necessary, and 2) sets up a basic shell hook that the active mise version can then hook into(?) I know, this probably sounds a lot easier than it actually is.
It's dead simple to manage files in git or such. What Chezmoi adds over that are niceties like:
* Permission management, so that ssh wouldn't refuse to log you in because ~/.ssh/authorized_keys is 0755 instead of 0644.
* Templating, because ~/.ssh/config has slightly different options on Mac and Linux, so you can't use the exact same file contents as-is on both systems.
I can run `chezmoi apply` and get all the files in the right places and they're all setup just right. Like so many others, I'd previously built my own ad-hoc system to handle these things, and it ended up looking like a crappy, half-baked version of Chezmoi. When it came up on my radar I immediately ported my own system over to it and never looked back.
Even just as a user of nix there has been this problem of how to manage dotfiles - people have git repositories for them but they are copies, because the actual dotfile in use is never tracked
For a System administrator the problem is many orders of magnitude worse
Which is pretty much exactly the same syntax you get with Hjem, but with more optional features. So you could just write that instead whenever you want full control :)
It works on darwin and any linux just fine. Theyre even working on a "standalone" like HMs, if you really want to decouple your user programs from your system.
With some programs it is as simple as you show, but with others there are options around the config and other nix module nasties that I avoid if I can. hjem-rum is a sister project to hjem that provides some similar modules.
I'm pretty committed to the `nix` ecosystem (I rewrote `nix` from scratch to unbreak it: https://gist.github.com/b7r6/90107d8e8ebe81fb0577b9c033b6ab0..., so, pretty committed), but I can't endorse it in it's current form after learning how the sausage is made and enough math to know why it's not just buggy but conceptually unsalvagable with the current `nixpkgs` and the current purity dogma.
And `home-manager` is maybe the most glaring instance of a tool that is demonstrably ill-posed where the "you're holding it wrong" from the community is a the community problem, not anyone holding anything wrong. From the Zed editor configuration stanza in `home-manager`: https://github.com/nix-community/home-manager/blob/a78606767.... That's not a Zed problem, that's a Nix problem. No one is holding it wrong, XDG config paths get mutated.
There is absolutely nothing "impure" about taking content-addressed bytes from a CAS (Xet in this instance) and surfacing them as a derivation. The "impurity" is called a "coeffect" and the action over the coeffect is called "grade discharge". This is thoroughly studied and works properly, it can cope with all of these cases and it's creates scope for dramatically more reproducible systems that are much easier to reason about (they are possible to reason about). Also, if you can't download shit from HuggingFace in 2026 without a weird hack where the name of the field is a scolding? That's gonna be putting downward pressure on adoption.
And most of the high-friction shit in Nix is like this, ignorance hardened into dogma hardened into theology. To wit:
- FHS vs. zany one-of-a-kind filesystems are nothing to do with purity or hermeticity or reproducibility, it's pure theology and the constant breakage with all the `patchFail` jank is at this point a jobs program, it's totally unnecessary. Namespaces/unshare, we have all the stuff to do this properly (`patchelf` and `unshare --bind-mount` are mathematically dual but only one blows content addressing).
- `drv` hash addressing is a reproducibility war crime. Floating CA is fine it's just broken upstream and in Determinate, it's not a valid ideological debate, it's bugs.
- there is absolutely no reason why the builder where a binary is produced needs to have the same filesystem layout or find libraries in the same place as the resulting artifact runs in any more than an adult needs to live in the same house they grew up in. `patchelf` works.
- having `libcuda.so.1` and friends at `/run/opengl-driver/lib` is dark comedy and source builds of `NCCL` when NVIDIA-certified binaries are in a wheel (zip file) on PyPI is the sequel. this is straight up bad for the planet and we should feel bad we haven't fixed it.
I could go on, but the main point for this thread is to the people who are on the fence about Nix: you're not holding it wrong the `nixpkgs` maintainers are holding it wrong, and more and more of us are getting serious about fixing it. Don't give up on declarative and reproducible systems that you can reason about because Nix is stuck in a weird place as software and as a community. There are reformers on the case.
These things really do not have an actual reason. Take for example the Nic people who go to every project with a bash script that uses #!/bin/bash and tells them to use #!/usr/bin/env bash. There is no justification for trying to fix every bash script in existence over the operating system being able to resolve what the "current" bash version should be (/usr/bin/env already has to pick a version). This is pointless O(n) busy work Nix people have created for themselves when an O(1) solution exists.
People shy from Nix because of supposed complexity but it really is the only real solution to this sort of problem. It's not really that much more difficult to learn, and in fact if you are willing, AI works really well generating nix config.
It took me a single afternoon to learn the basics and start using it. Contrary to what I initially thought, I didn't have to migrate all of my dotfiles at once. Then over the next couple of days, Codex migrated everything else for me.
One major benefit for me is that I no longer need to have once-in-a-while tools installed, because I can always spin up a temporary shell with `nix-shell -p packageName`. This significantly decreased the amount of software I have in my environment.
This works great with agentic coding. Agent wants to run `ripgrep`, but you don't have it? Tell it to run `nix run nixpkgs#ripgrep` instead.
But the biggest benefit is that now that you know Nix! So you can start using it to create reproducible development environments and uninstall mise, asdf, nvm, pyenv, etc. You can spin up reproducible servers running NixOS and never touch Ansible again. You can even install it in your router.
Or you can do none of that and continue just using it for your dotfiles. It plays nice with other tools.
It's not complexity its questionable documentation. Picking up Nix is really hard yet the best we got is a mishmash of unofficial recourses and many of them are out of date and/or focus on the packaging side which is terrible for introduction.
I spent a year dabbling in NixOS, getting a few "toy" setups.
About 6 weeks ago I installed NixOS on a spare laptop and did 100% of the configuration through Claude Codex. Initially I copied "/etc/nixos" off to my existing workstation, until Claude Code bootstrapped it enough to run on the NixOS machine.
I've been running it as my primary workstation for the last 3-4 weeks, and it's been great! 100% configured by Claude Code or Codex CLI.
Configuring a machine via Nix lang is not that fun. Configuring a machine via English is.
And I've thrown some curveballs at this setup. I asked it for gitbutler-cli, which NixOS doesn't/didn't provide, and it was able to package up and build it. It's running Sway. I have my secrets and configs in SOPS+Home Manager.
I am genuinely confused. Are you saying dotfile managers need to be more complex? Or that nix's problem space is the only solution for the things in nix's domain? (of which dotfiles are just one—via home manager)
I thought it was the former but "just pulling a precompiled, dynamically linked binary and hoping it works" makes me think we're not on the same page since to me that has nothing to do with dotfiles.
I'm saying that I value the completeness (thoroughness may be a better word) as well as the reproducability and portability that nix ensures over convenience. I cant tell you the amount of times I've pulled a precompiled tool and it just doesnt work because of one quirk of their packaging or another.
Also noting that I don't see the problem as "dotfile management" but as "system AND user configuration management" which extends beyond some plain text files in $HOME.
ed.: and home manager is just one tool which provides not only dotfile management, but drvs for installing particular programs and configuring then in highly opinionated ways -- I do not use it
> I can't tell you the amount of times I've pulled a precompiled tool and it just doesnt work because of one quirk of their packaging or another.
This is a serious problem in Nix too. I often trip over buggy and abandoned Nix packages. Nix often makes it easy to roll back or work around them, but they're absolutely there.
Are you really in this thread to talk about dotfile managers, or are you here to criticize Nix, given that Nix directly competes with your company's only product?
Considering nothing in that reply criticized nix you’re certainly wrong. I don’t hesitate to share my opinions on nix anytime (just like nix users themselves) so I’d have no reason to be coy about my criticism.
tfrancisl was polite (even though I arguably wasn’t in the beginning) so I felt it important to return the favor.
That problem is overstated. I've been dozens of custom packages and modules and it's not that much more complex that basic Nix config, and in fact is a huge positive, not negative. How easy can you write config that packages a binary and/or sets up a new service reproducibly? I'd much rather do that on Nix than Ubuntu or Arch for sure.
Once you hit the Chezmoi stage, you're only about 6 months from Nix and Home Manager. I mean, why climb _almost_ to the top of a mountain and then just sit down?
I've bounced off Nix every time I tried it, before I even started trying something like Home Manager.
I've been using (and contributing to) chezmoi for ~6 years now. Given that it has first-class integration with secrets managers, I suspect that it does things that Home Manager can't.
I ran NixOS for a while, before I switched to Apple Silicon, so I consider myself fairly well-versed-enough (although nowhere near an expert) in Nix and the Nix ecosystem. My last four jobs have all issued me MacBook Pros; the last three with Apple Silicon.
Ultimately, my workplace setup is what has the most gravity. And the most I can get most workplaces to standardize on is Homebrew for package management of off-the-shelf software. Nix is so far outside of the wheelhouse for most engineers that I can't even propose it. It would be too much of a distraction for too many people for too long that it's just not seen as worth it and it's not worth spending the political capital on the attempt. Employers would literally prefer to run scripts from a whitewashing, barely-auditable Jenkins instance with parameterized jobs than to attempt to figure out how to distribute portable scripts and get everyone's permissions working.
So I need to pick software that will cooperate with other tools in an unstable fashion, rather than software that attempts to fully and exclusively control the environment to provide guarantees. Chezmoi fits. Nix and home-manager do not.
You can run Nix and Home Manager on macOS or any Linux distro. You don't even need root.
It works exactly like you describe: it provides its guarantees for all software that you manage through Nix, and doesn't get in the way of software that you don't manage through Nix.
Unlike with NixOS, you can run binaries that expect an FHS-compliant system just fine.
You can just silently use it and enjoy the convenience for the declarative parts of your setup, with no detriment to your ability to run the imperative, ad-hoc setup scripts that your company requires.
> The trouble is that symlinks cut both ways. Every edit on every machine writes straight through the link into that machine’s clone of the repo
I have stow too in my micro DE, use it across a few machines and it's holding up really well. I designed my dotfiles so that changes would happens exclusively to files not tracked by stow. I have .zshrc tracked, but environment goes to .zshenv and general local customisation goes to .local/lib/zsh/overrides.zsh (https://gitlab.com/gabriel.chamon/archie/-/blob/main/deploym...). Hyprland config is tracked but device specific configs live in .config/hype/config/device.conf (https://gitlab.com/gabriel.chamon/archie/-/blob/main/deploym...). Deployment instructions/automation is to just copy dist folders in place, which aren't tracked by stow.
It depends on everyone's style. For me personally changes I don't remember having made to files I track in the repo are symptoms that I might need to review how I organize my files. Maybe those changes were something that broke and I forgot to commit after fixing. In any case for me this bidirectional nature of symlinks are a feature, not a bug in this management system.
I use syncthing to automatically sync my dotfiles git directory across PC/laptops, and stow to manually update symlinks when I add a new dotfile (the content of existing dotfiles is synced by syncthing already)
That way I don't have to remember to commit+push+pull changes to existing dotfiles (like bashrc or vimrc which I edit often) to sync them to other machines, it happens automatically in almost real time as soon as the file is saved on one of my machines (syncthing uses inotify to detect changes on monitored directories)
I always like to mention rcm. It's not a popular one but I tried probably 10 managers before it and much prefer it to the competition: https://github.com/thoughtbot/rcm
I looked at Stow and Chezmoi and also have stuck with YADM. The exact reason is that YADM is so simple and intuitive because it's basically Git-for-dotfiles with so little to learn. Yet it also manages to support alternate and template files.
It's great to manage your dotfiles, but I took it a step farther. I rebuilt the minimal Linux desktop environment of my dreams (startx, xinit, i3, i3status etc) with Ansible. It begins from a vanilla Ubuntu server 24.04.4 install. I bootstrapped it using a KVM + spice setup (using a spare physical SSD rather than a virtual one) and iterating over and over again until I finally got everything mostly working. I then booted off that physical disk, and kept iterating until everything was perfect. I've since adapted the setup to work on virtual aarch64 on macOS. I just recently tuned it to work on a crappy old Haswell Dell laptop, now properly detecting and configuring hardware vaapi capabilities, backlight, battery, trackpad, trackpoint, etc.
Pretty snazzy, watching YouTube in Firefox on a 13 year old laptop with hardware h264 decode and everything tuned exactly to my liking.
I tried GUIX a few times, but ultimately I couldn't quite get it working exactly the way I wanted it to work. I also didn't like the ugly filesystem layout that the store requires. I may get over it and revisit at some point. It will be a lot of work, but on the plus side I'll have a reason to learn scheme.
Switched to Chezmoi from random assortment of manually authored scripts. The workflow takes some getting used to, because I constantly edit the actual files without calling `chezmoi edit` first, and have to merge.
I like that when combined with `mise` (https://mise.jdx.dev) I can roll out a new computer in 2-3 commands and have my entire environment configured the way I like, with neovim and all the plugins and language servers.
I never did get used to the Chezmoi workflow. `chezmoi edit` never became muscle memory, and I was constant finding myself resolving diffs and what not.
I've never really understood why people get so fancy with their setups when you can just plop a git repo into your home directory. I suspect it has to do with people being unfamiliar with git.
For me it's useful to have a separate directory with my dotfiles repo, since my dotfiles repo's top level directory doesn't correspond to my home directory. I have one subdirectory for each machine.
I’ve always managed this problem in a different way. I don’t know if my way is better, but it works really well for me.
I treat my powerful desktop computer as my main machine. Then I have a bunch of laptops.
Then I just rsync my entire home directory out to all the laptops.
From there. The rule is quite simple. Any file created on a laptop are considered ephemeral. If I create data that I have to keep. It gets rsynced back the other direction to the main machine.
This process has served me well for at least 15 years now and is supported by a small handful of shell scripts to automate this process
Interesting. I go about this differently. I have one master setting and from there ruby just autogenerates anything I'd ever need on other computers. If ruby is unavailable then I just copy the generated files. But I only edit the master setting to enable what I need.
> This process has served me well for at least 15 years now and is supported by a small handful of shell scripts to automate this process
I feel in a similar way but not with shell scripts. Ruby autogenerates them if I need them too. Ruby is my ultimate glue to hold together everything.
Ok, I came into this thread intending to say “I’ve been using stow for years and am perfectly happy with it”, started RTFA and the comments, realised that I was actually not happy with it, started considering chezmoi then remembered that I had had a pretty great experience building a Nix VM recently.
Now I want to use Nix* to manage my multi-machine MacOS and Linux setup (with lots of dotfile config overlap, of course).
That’s the HN experience for you.
Kind souls: what is currently the blessed way to manage MacOS dots with Nix? I recall there is more than one paradigm - what’s the approach that simplest, most robust and can be adopted incrementally?
Edit: Just to say that I think Atuin now also plays in this space. Haven’t checked it out, though.
I must've tried to set up stow five or six times over the years, in between various hand rolled custom setups. I can't put my finger on why but I set up chez moi & it's been my setup since, much longer than any previous solution.
Chez moi is definitely not without its rough edges but it seems to have gotten the subtle essentials right enough for adhd me to not have abandoned it yet.
Same here. I would say chezmoi has almost a bit too many features. If one is into yak shaving there is a lot to explore. It's only a negative for me because I forget half of them everytime I read the docs. But I can hardly blame it for that, it's great!
I like to use environment modules [1] (or lmod [2]) for that purpose. You can make each manually built software package available or not on a per-shell-session basis, just by running
module load myapp
or
module unload myapp
in the shell where you want it (or don't). The small downside is that it only works with software that actually uses the standard environment variables like PATH, CPATH, etc. for their intended purposes rather than hardcoding filesystem paths, but in my experience it's rare to find something that doesn't. Also, you have to write a modulefile for each package, but that's not a big deal.
I learned about Stow after I found out about Chezmoi, and felt like Chezmoi was the better fit for me. I make heavy use of templating to keep work / personal aliases and functions separate, and could not be happier with the outcome.
Chezmoi strikes a nice balance between the overkill of home-manager while still being more powerful than simpler solutions.
Yadm is another alternative, the main thing I don't like about it though is that I'm not a fan of cross OS dotfiles. Having niri files on my work Mac and aerospace dotfiles on Linux annoys me quite a bit.
As powerful as the templating in chezmoi is, I think it should be considered a last resort and only used for simple files. They break your editor features like highlighting.
I hate . dirs. In fact, I hate them so much that I don't use them.
My configuration lives primarily in .yml files. These are kept
super-simple. When need be and another format is required, ruby
autogenerates these for me. For instance, all my bash aliases
are kept in .yml files which then get turned into bash rc files
or any other target format for other shells. Same for most of my
other configuration too - not always .yml but usually some text
file. I never understood the neet for .foobar directories or files.
They just hide a system that is intrinsically ugly and needlessly
complicated.
It’s quite new but I’ve been cooking up some new bootstrapping features with mise which people may find relevant here: https://mise.jdx.dev/bootstrap.html
It’s for things like dotfiles, apt/brew packages, and LaunchAgents/systemd.
EDIT: I feel a little bad having hijacked this, as someone that hears a lot of opinions about devtools I can definitely say chezmoi is a darling of the community and I highly recommend checking it out.
I am quite intrigued. With the sorry state of security, I am doing everything in VMs and have been trying to settle on the best way to setup a new machine. The process is so clunky that I end up defaulting to bigger instances than I should (more pets than cattle).
Being able to centralize this config is far more attractive than having a separate Ansible or pyinfra process.
Edit: The docs on this are very encouraging! However, it is not clear- what is experimental? Anything notably buggy or churning? Any far off features you hope to implement some day?
ALL mise features are experimental until the design is finalized. I don’t do breaking changes except in exceptional circumstances with a long procedure, so experimental means I have the liberty of making them in any release. It is not at all any marker of quality. That said, in practice I don’t make them often.
I’d argue it’s a good time to play with features since if you have any ideas on how it might better fit your workflow I still have the ability to change the design before it’s largely frozen.
i would reach for atomic fedora personally, maybe with bluebuild
That still feels like more external tooling, plus letting the distro make my packaging decisions. I am already using mise, so this is an almost free boon without more complexity.
I saw it the other day in the release notes and I'm definitely setting this up since I already use mise for many of my global tools and projects.
Some months ago I ported part of my git bare repo to chezmoi but never really picked it up since, even if I see it as a great and very complete tool for managing complex dotfiles, I don't have many machines and can live by with a bunch of if uname -s, and the fact that I still need to wrangle brewfiles and scripts for packages. Having the sytem package managers glue along the configs and symlinks is exactly what I wanted.
Just started moving things to mise and didn’t see this before, thanks.
Hopefully can use this alone instead of needing to combine w/ chezmoi / nix to get everything shell and pkg manager agnostic, consistent, and DRY (bash/zsh/fish + macports/pkgsrc/brew).
it's dependency free. You don't even need brew to install brew formulas.
The "exception" to that are linux package managers like apt-get and dnf which it calls under the hood. I think can't be an actual issue since it's not like you would ever use ubuntu/redhat without their system package manager installed.
Ha, I came here to share this! :)
Thanks so much for your work on mise! I used to be a heavy asdf user but nowadays I'm an even heavier mise user!
Random question while you're here: mise is undergoing pretty heavy development these days and I recently noticed that 1) my coworkers and I are not always on the same version, so some features/bug fixes are not available to everyone, and 2) package registries often don't have the latest mise version.
So I think we need a meta tool manager here to manage the tool manager version. :) Seriously, though, have you considered having mise manage its own version? I think that'd be pretty neat!
Thinking aloud, I guess one way to do this might be to distribute through package registries only a lightweight bootstrap application, which 1) reads the pinned mise version from mise.toml and downloads it as necessary, and 2) sets up a basic shell hook that the active mise version can then hook into(?) I know, this probably sounds a lot easier than it actually is.
The problem with mise managing its own version is perf. I don't want a shim that has to read config files to exec the right version.
I would make use of min_version. It's not perfect, but will at least help bring laggards along.
Are there plans to support MacPorts as a packaging system? I only use Homebrew for casks, because I find it unreliable for core development tools.
Haven't looked into it but agents are so good at this I bet it'll be trivial to add
I had similar problems with GNU Stow, but switched to Nix and Home Manager instead.
I think Chezmoi's templates and file naming conventions don't click for me, but it's nice to see a good variety in this problem space.
I didn't even know that managing dotfiles was a "problem space".
It's dead simple to manage files in git or such. What Chezmoi adds over that are niceties like:
* Permission management, so that ssh wouldn't refuse to log you in because ~/.ssh/authorized_keys is 0755 instead of 0644.
* Templating, because ~/.ssh/config has slightly different options on Mac and Linux, so you can't use the exact same file contents as-is on both systems.
I can run `chezmoi apply` and get all the files in the right places and they're all setup just right. Like so many others, I'd previously built my own ad-hoc system to handle these things, and it ended up looking like a crappy, half-baked version of Chezmoi. When it came up on my radar I immediately ported my own system over to it and never looked back.
Even just as a user of nix there has been this problem of how to manage dotfiles - people have git repositories for them but they are copies, because the actual dotfile in use is never tracked
For a System administrator the problem is many orders of magnitude worse
If you like home manager but also like to understand how your "home"/user programs are configured, take a look at hjem!
Last time I looked, it only worked with NixOS.
I don't understand what makes it more understandable than Home Manager, though.
If you look at the source of a Home Manager `programs.foo` module, it should look mostly like this:
Which is pretty much exactly the same syntax you get with Hjem, but with more optional features. So you could just write that instead whenever you want full control :)It works on darwin and any linux just fine. Theyre even working on a "standalone" like HMs, if you really want to decouple your user programs from your system.
With some programs it is as simple as you show, but with others there are options around the config and other nix module nasties that I avoid if I can. hjem-rum is a sister project to hjem that provides some similar modules.
I'll take a look again once it's a bit more mature. I need a standalone, single-user, rootless mode for it to work on all my systems.
Thanks for the recommendation.
I'm pretty committed to the `nix` ecosystem (I rewrote `nix` from scratch to unbreak it: https://gist.github.com/b7r6/90107d8e8ebe81fb0577b9c033b6ab0..., so, pretty committed), but I can't endorse it in it's current form after learning how the sausage is made and enough math to know why it's not just buggy but conceptually unsalvagable with the current `nixpkgs` and the current purity dogma.
And `home-manager` is maybe the most glaring instance of a tool that is demonstrably ill-posed where the "you're holding it wrong" from the community is a the community problem, not anyone holding anything wrong. From the Zed editor configuration stanza in `home-manager`: https://github.com/nix-community/home-manager/blob/a78606767.... That's not a Zed problem, that's a Nix problem. No one is holding it wrong, XDG config paths get mutated.
Another example and this is the one that really shows the shape of the thing: https://github.com/nixified-ai/flake/blob/bbd3a04fa1ae294096....
There is absolutely nothing "impure" about taking content-addressed bytes from a CAS (Xet in this instance) and surfacing them as a derivation. The "impurity" is called a "coeffect" and the action over the coeffect is called "grade discharge". This is thoroughly studied and works properly, it can cope with all of these cases and it's creates scope for dramatically more reproducible systems that are much easier to reason about (they are possible to reason about). Also, if you can't download shit from HuggingFace in 2026 without a weird hack where the name of the field is a scolding? That's gonna be putting downward pressure on adoption.
And most of the high-friction shit in Nix is like this, ignorance hardened into dogma hardened into theology. To wit:
- FHS vs. zany one-of-a-kind filesystems are nothing to do with purity or hermeticity or reproducibility, it's pure theology and the constant breakage with all the `patchFail` jank is at this point a jobs program, it's totally unnecessary. Namespaces/unshare, we have all the stuff to do this properly (`patchelf` and `unshare --bind-mount` are mathematically dual but only one blows content addressing).
- `drv` hash addressing is a reproducibility war crime. Floating CA is fine it's just broken upstream and in Determinate, it's not a valid ideological debate, it's bugs.
- there is absolutely no reason why the builder where a binary is produced needs to have the same filesystem layout or find libraries in the same place as the resulting artifact runs in any more than an adult needs to live in the same house they grew up in. `patchelf` works.
- having `libcuda.so.1` and friends at `/run/opengl-driver/lib` is dark comedy and source builds of `NCCL` when NVIDIA-certified binaries are in a wheel (zip file) on PyPI is the sequel. this is straight up bad for the planet and we should feel bad we haven't fixed it.
I could go on, but the main point for this thread is to the people who are on the fence about Nix: you're not holding it wrong the `nixpkgs` maintainers are holding it wrong, and more and more of us are getting serious about fixing it. Don't give up on declarative and reproducible systems that you can reason about because Nix is stuck in a weird place as software and as a community. There are reformers on the case.
What a snakepit of a comment! I know there are tensions within nix but this feels like a classic case of Chestertons fence at a big scale.
These things really do not have an actual reason. Take for example the Nic people who go to every project with a bash script that uses #!/bin/bash and tells them to use #!/usr/bin/env bash. There is no justification for trying to fix every bash script in existence over the operating system being able to resolve what the "current" bash version should be (/usr/bin/env already has to pick a version). This is pointless O(n) busy work Nix people have created for themselves when an O(1) solution exists.
Worse is better.
Mommy, the Nix Reformationist is scaring me
People shy from Nix because of supposed complexity but it really is the only real solution to this sort of problem. It's not really that much more difficult to learn, and in fact if you are willing, AI works really well generating nix config.
It took me a single afternoon to learn the basics and start using it. Contrary to what I initially thought, I didn't have to migrate all of my dotfiles at once. Then over the next couple of days, Codex migrated everything else for me.
One major benefit for me is that I no longer need to have once-in-a-while tools installed, because I can always spin up a temporary shell with `nix-shell -p packageName`. This significantly decreased the amount of software I have in my environment.
This works great with agentic coding. Agent wants to run `ripgrep`, but you don't have it? Tell it to run `nix run nixpkgs#ripgrep` instead.
But the biggest benefit is that now that you know Nix! So you can start using it to create reproducible development environments and uninstall mise, asdf, nvm, pyenv, etc. You can spin up reproducible servers running NixOS and never touch Ansible again. You can even install it in your router.
Or you can do none of that and continue just using it for your dotfiles. It plays nice with other tools.
It's not complexity its questionable documentation. Picking up Nix is really hard yet the best we got is a mishmash of unofficial recourses and many of them are out of date and/or focus on the packaging side which is terrible for introduction.
I spent a year dabbling in NixOS, getting a few "toy" setups.
About 6 weeks ago I installed NixOS on a spare laptop and did 100% of the configuration through Claude Codex. Initially I copied "/etc/nixos" off to my existing workstation, until Claude Code bootstrapped it enough to run on the NixOS machine.
I've been running it as my primary workstation for the last 3-4 weeks, and it's been great! 100% configured by Claude Code or Codex CLI.
Configuring a machine via Nix lang is not that fun. Configuring a machine via English is.
And I've thrown some curveballs at this setup. I asked it for gitbutler-cli, which NixOS doesn't/didn't provide, and it was able to package up and build it. It's running Sway. I have my secrets and configs in SOPS+Home Manager.
Guix solves the same problem in similar ways, though it uses Scheme as its configuration language.
Sadly it doesn't work on macOS, unlike Nix.
Or hardware.
It absolutely works with hardware. And it's not that difficult to add channels for extra software like a tainted Linux kernel.
And in return one gets a saner programming language than the Nix language.
how nix-pilled do you have to be to think that nix is the "only real solution" to dotfile managers?
Its the only complete, reproducible, and portable solution, in my opinion.
I like mise, but at the end of the day some programs are more complex than just pulling a precompiled, dynamically linked binary and hoping it works.
I am genuinely confused. Are you saying dotfile managers need to be more complex? Or that nix's problem space is the only solution for the things in nix's domain? (of which dotfiles are just one—via home manager)
I thought it was the former but "just pulling a precompiled, dynamically linked binary and hoping it works" makes me think we're not on the same page since to me that has nothing to do with dotfiles.
I'm saying that I value the completeness (thoroughness may be a better word) as well as the reproducability and portability that nix ensures over convenience. I cant tell you the amount of times I've pulled a precompiled tool and it just doesnt work because of one quirk of their packaging or another.
Also noting that I don't see the problem as "dotfile management" but as "system AND user configuration management" which extends beyond some plain text files in $HOME.
ed.: and home manager is just one tool which provides not only dotfile management, but drvs for installing particular programs and configuring then in highly opinionated ways -- I do not use it
> I can't tell you the amount of times I've pulled a precompiled tool and it just doesnt work because of one quirk of their packaging or another.
This is a serious problem in Nix too. I often trip over buggy and abandoned Nix packages. Nix often makes it easy to roll back or work around them, but they're absolutely there.
They still build though. No one said anything about Nix being a solution to buggy software or bad maintainers, that's an absurd expectation.
I don't think you're genuinely confused.
Are you really in this thread to talk about dotfile managers, or are you here to criticize Nix, given that Nix directly competes with your company's only product?
Considering nothing in that reply criticized nix you’re certainly wrong. I don’t hesitate to share my opinions on nix anytime (just like nix users themselves) so I’d have no reason to be coy about my criticism.
tfrancisl was polite (even though I arguably wasn’t in the beginning) so I felt it important to return the favor.
"real" meaning truly reproducible and not brittle.
Nix's complexity isn't with itself, its if you try to step one bit off the beaten path where it immediately starts to grate.
That problem is overstated. I've been dozens of custom packages and modules and it's not that much more complex that basic Nix config, and in fact is a huge positive, not negative. How easy can you write config that packages a binary and/or sets up a new service reproducibly? I'd much rather do that on Nix than Ubuntu or Arch for sure.
Once you hit the Chezmoi stage, you're only about 6 months from Nix and Home Manager. I mean, why climb _almost_ to the top of a mountain and then just sit down?
I've bounced off Nix every time I tried it, before I even started trying something like Home Manager.
I've been using (and contributing to) chezmoi for ~6 years now. Given that it has first-class integration with secrets managers, I suspect that it does things that Home Manager can't.
>I've bounced off Nix every time I tried it, before I even started trying something like Home Manager.
I did too. Until I tried configuring it with Claude Code. I'll give you my money back guarantee on it.
Vibed infrastructure? That sounds like it undoes a lot of Nix's advantages.
Or nix without home manager, and ideally without flakes as well. Two solutions looking for problems, IME.
I ran NixOS for a while, before I switched to Apple Silicon, so I consider myself fairly well-versed-enough (although nowhere near an expert) in Nix and the Nix ecosystem. My last four jobs have all issued me MacBook Pros; the last three with Apple Silicon.
Ultimately, my workplace setup is what has the most gravity. And the most I can get most workplaces to standardize on is Homebrew for package management of off-the-shelf software. Nix is so far outside of the wheelhouse for most engineers that I can't even propose it. It would be too much of a distraction for too many people for too long that it's just not seen as worth it and it's not worth spending the political capital on the attempt. Employers would literally prefer to run scripts from a whitewashing, barely-auditable Jenkins instance with parameterized jobs than to attempt to figure out how to distribute portable scripts and get everyone's permissions working.
So I need to pick software that will cooperate with other tools in an unstable fashion, rather than software that attempts to fully and exclusively control the environment to provide guarantees. Chezmoi fits. Nix and home-manager do not.
So don't propose it?
You can run Nix and Home Manager on macOS or any Linux distro. You don't even need root.
It works exactly like you describe: it provides its guarantees for all software that you manage through Nix, and doesn't get in the way of software that you don't manage through Nix.
Unlike with NixOS, you can run binaries that expect an FHS-compliant system just fine.
You can just silently use it and enjoy the convenience for the declarative parts of your setup, with no detriment to your ability to run the imperative, ad-hoc setup scripts that your company requires.
> The trouble is that symlinks cut both ways. Every edit on every machine writes straight through the link into that machine’s clone of the repo
I have stow too in my micro DE, use it across a few machines and it's holding up really well. I designed my dotfiles so that changes would happens exclusively to files not tracked by stow. I have .zshrc tracked, but environment goes to .zshenv and general local customisation goes to .local/lib/zsh/overrides.zsh (https://gitlab.com/gabriel.chamon/archie/-/blob/main/deploym...). Hyprland config is tracked but device specific configs live in .config/hype/config/device.conf (https://gitlab.com/gabriel.chamon/archie/-/blob/main/deploym...). Deployment instructions/automation is to just copy dist folders in place, which aren't tracked by stow.
It depends on everyone's style. For me personally changes I don't remember having made to files I track in the repo are symptoms that I might need to review how I organize my files. Maybe those changes were something that broke and I forgot to commit after fixing. In any case for me this bidirectional nature of symlinks are a feature, not a bug in this management system.
I use syncthing to automatically sync my dotfiles git directory across PC/laptops, and stow to manually update symlinks when I add a new dotfile (the content of existing dotfiles is synced by syncthing already)
That way I don't have to remember to commit+push+pull changes to existing dotfiles (like bashrc or vimrc which I edit often) to sync them to other machines, it happens automatically in almost real time as soon as the file is saved on one of my machines (syncthing uses inotify to detect changes on monitored directories)
I found chezmoi kind of annoying and then just moved to symlinks and a fossil repo. Low tech and low drama
I've been using [yadm](https://yadm.io/) instead which works really well!
I always like to mention rcm. It's not a popular one but I tried probably 10 managers before it and much prefer it to the competition: https://github.com/thoughtbot/rcm
I've been happy with yadm for few years now. I had tried chezmoi, but preferred yadm. I don't remember my exact reasons though...
I looked at Stow and Chezmoi and also have stuck with YADM. The exact reason is that YADM is so simple and intuitive because it's basically Git-for-dotfiles with so little to learn. Yet it also manages to support alternate and template files.
It's great to manage your dotfiles, but I took it a step farther. I rebuilt the minimal Linux desktop environment of my dreams (startx, xinit, i3, i3status etc) with Ansible. It begins from a vanilla Ubuntu server 24.04.4 install. I bootstrapped it using a KVM + spice setup (using a spare physical SSD rather than a virtual one) and iterating over and over again until I finally got everything mostly working. I then booted off that physical disk, and kept iterating until everything was perfect. I've since adapted the setup to work on virtual aarch64 on macOS. I just recently tuned it to work on a crappy old Haswell Dell laptop, now properly detecting and configuring hardware vaapi capabilities, backlight, battery, trackpad, trackpoint, etc.
Pretty snazzy, watching YouTube in Firefox on a 13 year old laptop with hardware h264 decode and everything tuned exactly to my liking.
don't want to be that guy, but have you tried nix?
I tried GUIX a few times, but ultimately I couldn't quite get it working exactly the way I wanted it to work. I also didn't like the ugly filesystem layout that the store requires. I may get over it and revisit at some point. It will be a lot of work, but on the plus side I'll have a reason to learn scheme.
Switched to Chezmoi from random assortment of manually authored scripts. The workflow takes some getting used to, because I constantly edit the actual files without calling `chezmoi edit` first, and have to merge.
I like that when combined with `mise` (https://mise.jdx.dev) I can roll out a new computer in 2-3 commands and have my entire environment configured the way I like, with neovim and all the plugins and language servers.
I never did get used to the Chezmoi workflow. `chezmoi edit` never became muscle memory, and I was constant finding myself resolving diffs and what not.
I made Janus which is kinda like a less bloated Chezmoi
https://github.com/logannc/janus
What was your inspiration for writing this? / What bloat did you shed relative to chezmoi?
I've never really understood why people get so fancy with their setups when you can just plop a git repo into your home directory. I suspect it has to do with people being unfamiliar with git.
Git works until you need conditional logic such as platform-specific files or templating.
For me it's useful to have a separate directory with my dotfiles repo, since my dotfiles repo's top level directory doesn't correspond to my home directory. I have one subdirectory for each machine.
I’ve always managed this problem in a different way. I don’t know if my way is better, but it works really well for me.
I treat my powerful desktop computer as my main machine. Then I have a bunch of laptops.
Then I just rsync my entire home directory out to all the laptops.
From there. The rule is quite simple. Any file created on a laptop are considered ephemeral. If I create data that I have to keep. It gets rsynced back the other direction to the main machine.
This process has served me well for at least 15 years now and is supported by a small handful of shell scripts to automate this process
Interesting. I go about this differently. I have one master setting and from there ruby just autogenerates anything I'd ever need on other computers. If ruby is unavailable then I just copy the generated files. But I only edit the master setting to enable what I need.
> This process has served me well for at least 15 years now and is supported by a small handful of shell scripts to automate this process
I feel in a similar way but not with shell scripts. Ruby autogenerates them if I need them too. Ruby is my ultimate glue to hold together everything.
Ok, I came into this thread intending to say “I’ve been using stow for years and am perfectly happy with it”, started RTFA and the comments, realised that I was actually not happy with it, started considering chezmoi then remembered that I had had a pretty great experience building a Nix VM recently.
Now I want to use Nix* to manage my multi-machine MacOS and Linux setup (with lots of dotfile config overlap, of course).
That’s the HN experience for you.
Kind souls: what is currently the blessed way to manage MacOS dots with Nix? I recall there is more than one paradigm - what’s the approach that simplest, most robust and can be adopted incrementally?
Edit: Just to say that I think Atuin now also plays in this space. Haven’t checked it out, though.
I must've tried to set up stow five or six times over the years, in between various hand rolled custom setups. I can't put my finger on why but I set up chez moi & it's been my setup since, much longer than any previous solution.
Chez moi is definitely not without its rough edges but it seems to have gotten the subtle essentials right enough for adhd me to not have abandoned it yet.
I used stow for a long time, then tried Ansible, but eventually settled into good old Make.
`make dotfiles` just creates a bunch of symlinks, takes 5 minutes, all good and happy. Everything is modular, declarative, simple. Never looked back.
Same here. I would say chezmoi has almost a bit too many features. If one is into yak shaving there is a lot to explore. It's only a negative for me because I forget half of them everytime I read the docs. But I can hardly blame it for that, it's great!
I nearly jumped from GNU Stow, but settled when I find the —-no-folding flag
i feel like using GNU stow to manage your dotfiles has always been a hack.. has it ever been a supported usecase?
stow is an indispensable tool for me to manage /usr/local for manually installed software. my workflow goes:
now, myapp and all its supporting files are in the right place in /usr/local. if i want to "uninstall", i just runI like to use environment modules [1] (or lmod [2]) for that purpose. You can make each manually built software package available or not on a per-shell-session basis, just by running
or in the shell where you want it (or don't). The small downside is that it only works with software that actually uses the standard environment variables like PATH, CPATH, etc. for their intended purposes rather than hardcoding filesystem paths, but in my experience it's rare to find something that doesn't. Also, you have to write a modulefile for each package, but that's not a big deal.[1]: https://modules.readthedocs.io/en/latest/ [2]: https://lmod.readthedocs.io/en/latest/
I learned about Stow after I found out about Chezmoi, and felt like Chezmoi was the better fit for me. I make heavy use of templating to keep work / personal aliases and functions separate, and could not be happier with the outcome.
~50 years of distributed systems research and this is a problem we still have to deal with today. Sad!
Chezmoi strikes a nice balance between the overkill of home-manager while still being more powerful than simpler solutions.
Yadm is another alternative, the main thing I don't like about it though is that I'm not a fan of cross OS dotfiles. Having niri files on my work Mac and aerospace dotfiles on Linux annoys me quite a bit.
As powerful as the templating in chezmoi is, I think it should be considered a last resort and only used for simple files. They break your editor features like highlighting.
I hate . dirs. In fact, I hate them so much that I don't use them.
My configuration lives primarily in .yml files. These are kept super-simple. When need be and another format is required, ruby autogenerates these for me. For instance, all my bash aliases are kept in .yml files which then get turned into bash rc files or any other target format for other shells. Same for most of my other configuration too - not always .yml but usually some text file. I never understood the neet for .foobar directories or files. They just hide a system that is intrinsically ugly and needlessly complicated.
You might like xdg-ninja‡ then, it tells you what program put which item in your $HOME and how to move it to ~/.config or $XDG_CONFIG_HOME.
‡: https://github.com/b3nj5m1n/xdg-ninja
Found the Kubernetes guy
;)