I’ve been reading up on the concept of dotfile management, and I’ve come across tools such as GNU Stow, chezmoi, and yadm, but before I select a tool, I’d like to understand the workflow a little better.
I understand if the dotfiles are in some cloud provider such as GitHub, then after a fresh install one can do git clone etc, but let’s say one’s dotfiles are not stored in the cloud, then what’s the workflow for getting those dotfiles onto the freshly installed OS? Do people do git clone from another machine on their local network, manually copy the dotfiles folder from the source, use an app like LocalSend, or something else?
EDIT: Clarifying that this is for a home environment, where I have two or three different laptops in service at any given time. One is my main daily driver and doesn’t change much. The other two are kinda my sandboxes and I’m often distro hopping on them. This question is mostly for distro hopping on my sandboxes, which is like once or twice a month. Thanks!
You could self-host a shared “source of truth” git repo that you access over ssh or filesystem. That can be anything from a USB thumb drive, a small clean server or a container on your existing desktop with ssh access, to an entire Forgejo deployment. Then you only need the “secret zero” of an ssh key to get everything set up and syncable.
If fresh setup is more common, you probably have other parts like package installation and network configuration that you also want to automate. Enter configuration management like ansible or salt, image builders like packer or archiso, “immutable” solutions like Nix or rpm-ostree. Once you get there you typically manage that in git anyway and you could put your dotfiles repo as a submodule and copy them over as part of OS setup.
If it’s just for once in a blue moon, manual ad-hoc copying gets you pretty far.
No matter how you slice it I think you have to either frequently spend time syncing changes or just accept the drift and divergence between machines and the sources.
I made a script just for this purpose, I run the script on a fresh system and it pulls my stow directory without me needing to manually mess with ssh keys or passwords.
On a flashdrive, I have a folder named “setup”. In that folder, I have this script called “run” and a directory called “ssh”. In that “ssh” folder (not to be confused with ~/.ssh), I put my private ssh keys and their pubs.
#!/bin/bash # stop script immediately on error set -e # change working directory to directory containing this script cd "$(dirname "$0")" # check that ./ssh exists and exit if not if [ ! -d ./ssh ]; then echo "./ssh not detected, exiting..." exit 1 fi # create .ssh directory [ ! -d $HOME/.ssh ] && mkdir $HOME/.ssh chmod 700 $HOME/.ssh # copy keys to ~/.ssh cp -a ./.ssh/. $HOME/.ssh/ # ensure right permissions for .ssh contents # note: 2>/dev/null suppresses errors if no .pub files exist, || true to avoid exiting on failure chmod 600 $HOME/.ssh/* chmod 644 $HOME/.ssh/*.pub 2>/dev/null || true # start ssh agent eval `ssh-agent -s` trap "ssh-agent -k" EXIT # add keys ssh-add "$HOME/.ssh/privatesshkey" # add known hosts # note: removing them first then adding again to avoid duplicate entries ssh-keygen -R codeberg.org 2>/dev/null || true ssh-keygen -R github.com 2>/dev/null || true ssh-keyscan -H codeberg.org >> $HOME/.ssh/known_hosts ssh-keyscan -H github.com >> $HOME/.ssh/known_hosts # clone repo cd $HOME if [ -d "$HOME/stow" ]; then TIMESTAMP=$(date +"%Y%m%d_%H%M%S") mv "$HOME/stow" "$HOME/stow.old.$TIMESTAMP" fi git clone ssh://git@gitprovider.domain/myusername/stow.git