Building your own desktop environment

Updated 2017-05-01

All-in-one desktop environments (DE) such as Gnome, KDE and those found in popular systems get in the way of the power user in that they do not leave much room for customization, it restricts the potential a computer can provide. Moreover most popular DE have been plagued for decades with unpractical, non-automated and limited window management.

Positioning and resizing windows does not take much time per se, but it stacks up. Think about how many windows you fire up every day (which, for some obscure reasons, seem to be spawned at random places all the time on many popular systems…). Add up all the seconds you spend managing them, moving around, resizing… Wouldn’t it be nice if this were done for us? After all, a computer is an augmented automata, it should be able to automate this.

Another side-effect: the classic desktop environment (DE) will bundle more than you want… As such, they will always require more disk space and virtual memory than you actually need. The following advises will make things lighter and faster.

Dynamic Tiling Window Managers

There are ways to automate window management: tiling window managers. Once we have gotten used to them, it gets hard to bear with the waste of time incurred by the “regular, user-friendly” window managers. (Like those featured in Gnome and KDE.) The latter are called stacking window managers. I find the concept of stacking fairly questionable: why would the user want to cover part of other windows? Either you want to have a full look at some of the currently opened windows, or you just want to focus on one windows. Which means either tile the windows or go fullscreen. In other words, unused screen space is wasted screen space. I know how badly we like to look at our wallpaper, but that is not the primary use we should make of our computer.

This leads us to the antagonists of the stacking window managers: the tiling window managers. Windows do not “float” anymore, i.e. their position and size is not controlled manually but automatically so that it fits the screen space best. Understand that tiling managers do no deprive you from manual control over the windows: they are, by definition, a superset of stacking WMs, and thus you can still resize and position windows as you see fit.

The core idea is to be fast and display what the user wants.

See how mobile devices work: the restricted screen space does not allow for stacking windows, so all applications go fullscreen.

TWMs allow the user to relay full window management to the WM independently of the application capabilities. For instance a good WM will let you go fullscreen or use tabs with any program. This is part of the Unix philosophy: instead of duplicating functionalities across programs, write it once and for all where it is meant to be. File browsers and Internet browsers should not have tabs: WMs will do a better job (grouping, sorting) while keeping consistent behaviour and key bindings across programs.

Last but not least, there is a subcategory among tiling WMs: the dynamic ones. The tiling is done whenever a window is fired up.

In theory, the concept of a window manager is rather simple. In practice, they mainly differ by their configuration. Some WM will offer little tweaking possibilities to the user, while other will allow programmable modifications. The former is easier to get started with, while the latter will allow for the exact desired behaviour.

In my opinion, the most powerful WM would be a DTWM with a Turing-complete configuration. Such popular DTWMs include Awesome and i3. Awesome’s configuration is in Lua, i.e. it allows virtually unlimited tweaking, while i3 relies on external shell scripts.

Keyboard control

The use of the mouse has become so widespread with the advent of desktop computers that it now seems “intuitive” to associate the use of a mouse together with desktop environments. All the desktop management (windows, firing up programs, monitoring the system, etc.) are discrete actions, they do not require the continuous movement of a mouse pointer. As such, everything can be bound to a keybinding. Go fullscreen? Keybinding. Change the window layout? Keybinding.

Keep in mind we say “keyboard shortcut”: the word shortcut is there for a reason. See my article on mastering the keyboard.

Building from bricks

Refraining from using a full-featured desktop environment means that you will have to install the components one by one and configure it all to your liking. Which is the only way to fully tailor a computer to your needs. While it may seem daunting at first, this is a one-time task: it is straightforward to re-use the configuration across systems when it is stored in text files.

Text configuration

Most desktop environments come with a configuration panel: a program that is meant to centralize configuration. In practice, this is an additional layer of complexity that is not really needed.

Editing a text file has many advantages:

This last point is key: it allows you to watch over the evolution of your configuration changes (diffs); it also lets you deploy and sync a configuration across several computers in an instant. Last but not least, it lets you share the configuration!

Independent power-apps

Desktop environments push for developing “desktop-specific” applications, such as the K* programs for KDE vs. the G* programs for Gnome. I find this very embarrassing: why would we spend months and years developing a program for a sub-group of people? Why not spreading it to the world? Is it a form of geek-racism?

Our favorite programs should not have to do with the desktop environment. If the latter goes out of fashion and gets unmaintained, does it mean we are supposed to change the whole set of our programs, possibly losing some functionalities in the process?

A “desktop” in my sense is just the WM with a launcher. The rest should be the user favorite programs: a terminal emulator, a desktop monitor, a text editor, etc. Every feature from a desktop environment is a program. Missing this volume icon? Add key bindings. Want to tweak the network configuration? Fire-up the appropriate tool with a keybinding, or go edit the text file full of endless possibilities…

See my list of power-user applications for some inspiration.

Black & White

When the mere mortals have a glimpse at my computer, I would expect their first reaction to be “What system is it?” or “Why is it so barren?”. No, it is invariably “Maaaah, why is it all black?!?”

I find it interesting that black has become shocking. Black used to be the standard back in the days. The white-dominant background may have been popularized by the “user-friendliness” marketing of Apple and Microsoft in the 1980s as a reaction against the “hard-to-use” Unix and friends: the color was here to mark the difference. The association of ideas was easily made: white is brightness, it is light, as opposed to the uncomfortable, insecure darkness of Mr. Black.

Now if we are rational, it is just a color. But there is more to it. Working on a dark background will not strain the eye as much. The whole industry is warning its professional, intensive users against the possible eye strains and headaches resulting from staring at a screen for too many hours.

Solutions are developed: take a break, use eye drops, and whatnot. While it would be so much easier to eliminate the problem in the first place… Listen to the wisdom of our forefathers, they were bright people! Working on a dark background will solve the problems all at once. To the point that once you get used to it, switching to a bright background will stun you on the spot!