Miracle WM, a new tiling window manager built on Mir

Version 0.1.0 of Miracle-wm is very incomplete still, yet shows that interesting stuff is happening on the back of Canonical's Mir display server.

Canonical developer Matthew Kosarek has released version 0.1.0 of his project Miracle-wm, a new tiling Wayland compositor. There are many such tools around now, but a couple of things are unusual about this one: the display server it uses, and its packaging.

So far, this is just the first incomplete preview release of Miracle, and Kosarek has also published a roadmap detailing where he hopes the project will go next. He's aiming for two more preview releases before a complete version 1.0 in mid-July.

Miracle-WM is a new Mir-based tiling Wayland compositor, shown here with Waybar for added decoration. - Click to enlarge

We tried it, and it does what it says on the tin, which is to say very little at all just yet: you can open windows, and it will tile them for you – via customizable keystrokes – and that's about it. The sample configuration file in the user guide suggests partnering it with helpers such as swaybg for wallpapers and Waybar as a panel. (We found that also installing fonts-freefont-xyz will get most of Waybar's indicators to show properly.)

• Starting over: Rebooting the OS stack for fun and profit
• Drowning in code: The ever-growing problem of ever-growing codebases
• Latest version of Canonical's Wayland compositor arrives
• Fresh versions of Ubuntu Touch, Mir display server, and Unity arrive

This isn't a criticism: tiling window managers tend to be minimal by design. The jaded old vulture behind the Reg FOSS desk suspects this is due to the fact that a lot of Linux users tend to be keyboard warriors who do a lot of their computing in a collection of terminal-emulator windows, alongside a web browser window or two. Many hardcore shell-pilots regard things like file managers and desktop icons as unnecessary fripperies.

Tiling window managers have been growing in popularity in recent years. As a guideline, the Arch wiki lists no fewer than 14. That's just traditional X11-based ones – and it must follow, that where there's been lots of development activity around X.org, you will also find lots over on the Wayland side of the fence, where by comparison Arch lists a baker's dozen. Even GNOME is considering it.

What's interesting about Miracle is that rather than being either a standalone compositor in its own right, or being based on the wlroots compositor like many of the existing ones, Miracle is based around Canonical's Mir display server instead.

Wayland isn't a display server; it's just a protocol. That means that every Wayland-based environment has to implement its own program, called a compositor, to do the job of what was historically a separate tool in its own right: a display server. For 25 years or so, that meant X.org or its forerunner XFree86. So, in GNOME it's Mutter, in KDE Plasma it's Kwin, and so on. Mir goes against this trend: it's not especially tied to any particular environment, and as we wrote last year elements of multiple different environments can run on top of Mir. As such, we're happy to see tools built upon it. Mir has the potential to act as a unifying influence in this domain.

The other interesting aspect about Miracle is how easy it is to try, and indeed, to remove again. Being a native Ubuntu program from a Canonical developer, it's packaged as a Snap. Installing it is as easy as:

sudo snap install miracle-wm --classic

Reboot, and it appears as an option on the login screen. This shows the way that Snap, as a format, is more capable than Flatpak, which struggles with such low-level system components and has a notably clunky way to run command-line apps. In the project's README, Kosarek says:

While the project is only built as a snap at this moment, I am not allergic to other packaging formats, just perhaps too lazy to implement them at this moment. I will happily accept contributions in this domain.

Bootnote

We cynically suspect that a second reason for the proliferation of window-tiling systems is widespread ignorance about the standard keystrokes for controlling Microsoft Windows, which also work fine in the better-executed Windows-inspired desktops out there. If you've spent time and effort learning the control keystrokes for an obscure Linux text editor, then it's perfectly reasonable to want to use similar shortcuts to control the layout of your windows – or be able to construct your own set.

The fact that there are already a pretty complete set of such keystrokes and have been for about 35 or 40 years isn't widely known. Not knowing that the wheel exists is strong motivation to re-invent it. ®