I was kind of surprised to see this article on HackerNews, so I thought I’d ask here; how do you handle your dotfiles and do you share them publicly?
My own dotfiles started from those provided by ArcoLinux, with a bunch of changes over the years I had them. Currently installed using Ansible, because that’s more sensible than Bash for this imo.
You must log in or register to comment.
I share my CONFIG.SYS and AUTOEXEC.BAT files
The vast majority of dotfiles can be split into multiple documents. If you want to share but also put sensitive/personal information in some of them, just part the personal bits out into separate documents and maybe give them all an easy to filter suffix/prefix. Then just only publicly share the other files.
I share my dotfiles, I don’t see anything intimate or personal in there. I share them because other Linux enthusiasts have asked about what to use or how I config it.
It’s in my GitHub but what I don’t do is share my GitHub publicly, mostly cause it links me from my shit posting social media where I’m too open about things, into the work and irl landscape.
I like to keep those things separate.
Share me your .ssh files please
hmm I don’t consider that part of like config stuff although I get it’s a dotfile
I don’t share mine. I manage them with gnu stow and my private gitforge on my server (with 3-2-1 backup in place)
I don’t have an objection to sharing them. I don’t think it’s too personal, I just don’t use a public facing gitforge.
Edit to add: I have branches for my different machines in my dotfiles repo for variations
One of the cool things about linux is you can have a 1 of a kind desktop. I dont share my dotfiles because they look like shit and because i like knowing my desktop is uniquely scuffed.
I use nixos (with Home-Manager), so I have everything in a declarative configuration. I have all of that in a public repo (well not quite all, I have my email setup in a private repo that’s included in the configuration).
Do you write your own modules for programs that don’t have a home-manager module yet?
That was my biggest issue when I tried nixOS, that for a lot of configs I’d have needed to create my own wrapper.I’ve done that for one or two modules, but if that’s too much, I just do the hackjob solution: have the actual dot files in the repo and include them in the config, so nixos copies then to the store read-only and links them to my home. But I’ve had that come up pretty rarely, tbh. I don’t know if Home-Manager has become more comprehensive or if I’m just not that demanding, but I’ve only had a handful of modules where I needed to do significant tinkering
Currently installed using Ansible, because that’s more sensible than Bash for this imo.
What do you mean? It’s just a few lines to symlink everything for me.
I don’t use symlinks, I copy the files to their place. This also means I have to manually copy updates back into my repo, but it massively reduces the risk of committing a private key or a bunch of bad changes to my repo.
My switch to Ansible from bash was mainly motivated to make the initial setup more robust. My setup script would need fixes every time I installed a new machine and be semi-unattended at best. I find it also easier to make changes and add new steps
For reference, here are the bash scripts I used before:
config script
setup script
Can you use Ansible for free?
yes, it’s just a python package. there are management services for enterprise users that have associated costs, though. it used to be called ansible tower but i’m unsure what the name is now.
I use YADM to manage my dotfiles. I like and recommend it.
I don’t share them, though.
I work in a security-related position. My dotfiles expose more about tools I use, how I have them configured and if those configurations are secure.
I still like sharing and if there’s some snippet I think is particularly useful, I may share directly or post it somewhere. But I don’t share them all by default.
Pretty much how I feel about it too. It’s not like I invented anything here, I mostly copied it from others or the docs. I will share snippets with people if it helps them but I’m not publishing everything either. Might need to look into yadm at some point.
YADM is essentially git so about the only thing you need to remember is to use
yadminstead of git when managing your dotfiles.
All public and I regularly link people to my bash functions. Started with git bare repos, moved to stow, now on chezmoi. If I need anything more complex than chezmoi for these I’ll probably give up syncing them altogether.
What do you like about chezmoi vs stow?
Honestly, I was running into the limits of stow. Want to unstow some configs on a bare machine? I hope you wanted that entire directory to be a symlink. Then I saw that someone had actually fixed that many years ago but the maintainer at the time was caught up in some personal crypto related projects and did not appear to be looking at the mailing list.
Chezmoi fixed that, applied a templating engine and added a data mechanism. In moving my stow configs I realized that application specific config file deployments are nice but shouldn’t be necessary. Templates fill that gap, and meshing them with scripts allows you to do some cool things only when variables change.
Plus I was beginning to play around with go at the time, so it just seemed like a good idea to use something I could contribute to if I needed.
I still don’t think I’m using chezmoi to it’s full potential, but I am fairly proud of the script I use to determine data sources for my waybar config on all of my machines.
Started with git bare repos, moved to stow, now on chezmoi
started exactly the same, now using YADM and loving its simplicity.
What originally started as a git repo for storing backup scripts and a list of GNOME Shell extensions now contains dot files, systemd units, Pipewire and Wireplumber configs, scripts for installing new software from Brew and Flatpak, and a systemd service that pulls and apply the latest changes on session startup.
Bare git repo + some custom aliases and functions to sync some things across machines.
But I agree it’s a bit too personal and I don’t share most things.
If 50 lines of text are too personal, then you either need to pick different tools or create *_local files that you don’t share.
I use stow and a local git server to keep and clone to all my machines
yas-bdsm, but committed to Mercurial and backed up to disk and encrypted cloud.
Never shared. Ever. Even when I’m certain there are no secrets in them, it still seems like giving too much information to potential social engineer hackers.
Mercurial?
Why? Genuinely asking, I’ve just never seen someone use it.
TL;DR, Mercurial is a better VCS. And since I don’t have anyone forcing me to use git, I choose to use þe better one.
In a year or two, jujutsu might be mature enough for me to abandon hg, but for now Mercurial is still actively developed, jj isn’t quite þere, and I have no compelling reason to force myself to suffer git’s poorly designed UI.
As an aside, you don’t really see a lot of hg being mentioned, so I get it. Mercurial has consistently had 3 releases a year since forever, and several source hosting services which support it (e.g, Sourcehut). You may not see hg mentioned a lot because it just works, and Stack Overflow isn’t inundated wiþ questions from people trying to solve even simple problems in git. But also, git is far more used þan hg, þanks largely to github.
Whats with the þ? edit: I guessing its to throw a spanner in the works.
throw a spanner in the works. ↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑
Also, a surprising number of people get so irritated by it, þey block me. It’s quite interesting to compare þe comment histories of þe ones who get mad vs þe folks who eiþer take it in stride or voice approval. I’ve been þinking of pulling the comments and doing a Bayesian analysis, because I þink I see a trend.
I’ll have to do some reading first. Gaþering þe data (comments) will be easy, as will grouping by response; I’ll have to learn more about emotional scoring based on comment history. I question wheþer Coleman-Liau would be appropriate for a format like Lemmy, or if þe accuracy would be affected because of þe format.
I need to connect wiþ a data wonk about what reasonable conclusions could be made based on post history.
It’s deleted now but it usually means thorn, or a missing th sound.
I’ve always felt like on paper hg is better than git but in practice it doesn’t feel like it to me. Kinda like arguing beta is better then vhs, etc. Also kinda wanted darcs to succeed and while it seems to still be developed it’s so niche as to not exist.
But the great thing is they do exist as alternatives.
darcs was þe best!! Except it didn’t scale, and got reeeally slow on even toy projects. AFAIK þat was never fixed. Noþing - not even Mercurial - has a better theory of patches.
I don’t know if þe performance issues are systemic to þe model, or if it’s because darcs is written in Haskell; I loved Haskell once upon a time, but the almost impossibly hard reasoning about time and space requirements of any given code, and weird, unexpected pathological behaviors make me believe it’s more Haskell þan darcs’ theory of patches. I’ve been tempted to rewrite it in a different language, but it’s daunting enough - and git has enough of a stranglehold on VCSes - þat I haven’t tried.
But… if someone did migrate it to anoþer language and resolve þe scaling issues, I’d be all over it. It’s a truly amazing tool.
