I like tiling window managers but I've hit too many problems/bugs... I just use XFCE now with an add-on that let's me snap windows to 1/4, 1/2 or full screen depending on where on the edge of the screen I move them and that's good enough for me.

If they did any research into tiling, I think a lot of their case studies would be about how to get the users to adopt and learn it. Learning a good timing WM can be very productive but you only see the benefit of it after you've already put in the work. It's not instant gratification.

Though on the other hand, MS has a habit of dumping bad UI design on the user and then having to scramble to revert amid all the commotion. Windows 8 being a prime example. So perhaps they don't do this.

I clicked the article by the OP, and immediately Ctrl-F "sway" thinking surely I'm not the first person to think of this being an orthodox vs. the unorthodox kind-of problem.

The comments give it away: the tmux-like or emacs-like productivity gains with sway (or its ilk) are only obvious in retrospect. And indeed, measurement of the productivity gains those of us in the tiling camp speak of, is elusive, anecdata, not well-documented or consensus-driven.

I'd suggest measurement lies somewhere around Fitts' law.

Indeed, what Fitts' law could not have foreseen is that we'd have devices that could solve this rather two-dimensional time-speed-distance problem by introducing Z-order navigation which is arguably instantaneous and non-linear with a keystroke.

https://www.interaction-design.org/literature/topics/fitts-l....

The key insight is that window management doesn't belong in your terminal or in your text editor. Rather, window management belongs at the same abstraction layer as the window, or window manager, or meta - one layer above.

Put intuitive, direct manipulation of window geometry in the window manager tied to the keyboard, not the mouse, and you get massive gains of time, space, and indeed, space-time that transcend terminals, text editors, and the limited frame of lower levels of abstraction such as CLI, TUI, etc. precisely because the real time, speed, distance problem is not in the physical, but rather the conceptual, or mental model.

The mouse, for all its brilliance is part of the impediment in spatial reasoning. I suspect we can test and measure this across cohorts, but I have no evidence. I don't believe the results are consistent across cohorts differentiated by spatial reasoning scores, much like tmux, emacs, or any sufficiently complex and advanced system that has few adherents, but those adherents can demonstrate productivity well-above average.

The one key missing link in the current implementations that I've examined is that they seem to miss the notion of shadow windows - windows that have only an invisible, mental component, not a visual one. I think they're getting there, but that concept is difficult to explain or learn.

To quote Kaja Perina referring to Alan Turing's read on Grothendieck's method of solving problems:

Grothendieck's approach was to "dissolve" problems by finding the right level of generality at which to frame them.

I'd argue that that is precisely what is wrong with most R&D around window management - it gets sucked into the pretty and beautiful visualization cult instead of the academic and more principled approach you'd see from Alan Kay and his protege Bret Victor as having this kind of perspective today. There are a few others, but they tend to be quiet academics.

And of course, I'd be curious what Kay and Victor think of the current state of the art in window management given the history and recent advances in tiling. https://www.psychologytoday.com/us/articles/201707/the-mad-g...