A few days back I've been describing a system to deploy/withdraw personal configuration files called `dewi'. The system actually works. I'm using it for the vast majority of my configuration files.

When I want to get my zsh config to a machine, I'm "git-cloning" the repository and call "make deploy" in it (assuming dewi is properly set up). But when I get access to a machine, I typically want more things set up than just my shell. I could do all that by hand. I could. But I'm much too lazy for that. So, I hacked up two scripts that take me from zero to hero with just a few commands. I'm calling that procedure bootstrapping an account.

The following assumes that `dewi' is installed on the host system. Also, `git' is required to clone some repositories.

Say, I got my user `ft' set up on a system. Its entire home-directory is empty (except for the system's skel stuff, which happens to be some minimal bash setup). Let's go:

% mkdir ~/etc
% cd ~/etc
% git clone git://github.com/ft/_bootstrap.git
% cd _bootstrap
% ./bootstrap.sh yes_please
% ./get_repos.pl 0x50.de _all_
% cd ..
% make -i deploy
% chsh -s /bin/zsh
% logout

And that's it. Next time I login, everything will be setup the way I'd like it to. That's five commands. Ten if you count the chdirs and the logout. If you can do this quicker, I'd love to see how. :-)

The interesting lines in the above list are "./bootstrap.sh yes_please" and "./get_repos.pl 0x50.de _all_", because they do the actual bootstrapping. The "make -i deploy" is just dewi deploying everything from `~/etc'.

What does bootstrap.sh do? Well, it calls `dewi'. It initialises `~/etc' as my dewi-root-directory. Then it takes some files files from the `addons' and `cfg' directories of the `_bootstrap.git' repository and deploys them into `~/etc/.dewi' using `dewi' itself. After that the system is completely usable with all my customisations on board.

Then, `get_repos.pl' comes into play. It's just a flexible way to clone git repositories that contain my setup files from somewhere. (Oh yeah, sidenote: because of the nice guys at dev.0x50.de, I got a place for my setup files, accessible from virtually everywhere: dev.0x50.de/projects/ftdotfiles666 If you're reading this, thanks a lot again!)

That should also explain the `0x50.de'. That's what I call the source. It's defined in a file in the `sources' directory. It maps configuration files names to a repository. The `_all_' parameter is special. It tells the script to just fetch all the repositories from the named source. Usually, you'd specify one of more categories as defined by the files in the `lists' directory. So, if you'd like to fetch the `core' and `editors' config repositories, you'd do:

% ./get_repos.pl 0x50.de core editors

I made two special categories for types of machines, too. `remote' and `windoze'. The former is for machines I got remote access to, the other is for Microsoft Windows 7 machines, with which I'm stuck with at work. So, when I get remote access to a new machine, I'm following the above listing and drop in:

% ./get_repos.pl 0x50.de remote

...and be done with it (on Windows 7 machines that's "./get_repos.pl 0x50.de windoze" from a cygwin prompt).