Xmonad seeking help for Wayland port (2023)

That is the thing with Wayland, it is much harder to create a window manager for Wayland. IIRC, fvwm decided not to create a Wayland version due to the difficulty.

When Wayland replacing X, lots of cool window managers and mini applications will be gone.

There are libraries like wlroots (C) and Smithay (Rust) to be able to more easily create your own wayland compositor

Are you suggesting every application should implement its own compositor?

The blame could be as well on Haskell.

IMO Ganeti died because of such a choice. There aren't enough programmers that are willing to invest into that niche.

I have nothing against that language per se, still such a choice can easily develop into a dead end.

This style of engagement is so off-putting.

This isn't wayland's fault. It's the compositor implementing wayland's fault for not exposing a window manager API. Nothing about wayland prohibits the creation of a window manager API.

What about wayback? Assuming running X by itself becomes real bad and undesirable, would wayback+Xwayland cover all those "can't Wayland" use cases? What remains (besides better stability and wider availability of wayback) to be done?

However the radically different architecture of Wayland may necessitate a rewrite well beyond what the maintainers of a window manager feel is easy.

Even accounting for wlroots, you're not exactly just running sed on a glob. And unfortunately, wayland didn't actually fix X's complexity problem. Arcan did, but we're not allowed to have nice things because Redhat has no taste.

I’ve heard the same thing about Wayland and NVidia’s drivers. To me, it seems like Wayland was designed to push all the hard work onto everybody else. That way Wayland never gets blamed for anything!

Yeah, I’ve long had fantasies of writing a little desktop for myself next time I get a long stretch of time off, but that became much more daunting with the advent of Wayland, even when factoring in the existence of wlroots and such. It’s like going from building a bicycle to building a modern fuel injected car with an automatic transmission.

There are multiple window managers in other languages that won't build a Wayland equivalent due to the effort so it's not only about the language

We also got a lot of new little niche window managers. hyprland, niri, cosmic, sway, river, labwc, dwl, wayfire and vivarium which is xnomad inspired...

For practical purposes, the problem with Wayland from the WM-dev's PoV is that you're either implementing a huge project or you're depending on wlroots, and wlroots still isn't where it would need to be for implementing a simple window manager to be as easy as it was with X11.

From the Wayland devs' PoV, mainstreaming Wayland successfully shifted responsibility for doing most of the heavy lifting in the graphics layer from the neglected X-Windows project to the well-established KDE and GNOME. The state of wlroots and the ecosystem of personal WM projects is unavoidable collateral damage.

For an individual developer, perhaps the thing to do is take a page out of bbLean's [1] bag of tricks and implement your WM on top of one of the big two desktop environments.

https://bb4win.sourceforge.net/bblean/

Not having a defacto compositor was a major blunder and resulted in an enormous delay to the project, reputational damage, and numerous challenges for app developers.

The problem is that compositors aren't giving you an API to target. Hyperland has plugins, but that is a whole other can of worms.

The libraries help, but it's still a bigger job than a window manager was

What I don’t get is why they pushed the compositor onto WM developers in the first place. Compositing seems like way too low level of a task for a window manager to be concerned with.

Even still, they are very hard to work with when compared to X11.

Fvwm people are very smart and have been developing fvwm for longer than Linux have been around. From what I understand and have read, bring fvwm or creating a fvwm clone on Wayland is near impossible. Far too much work.

So we will really end up with "pigs" like Gnome3, KDE or a slew of tiling environments. None of the cool WMs like Windowmaker, fvwm, dluxbox, twm, ctwm, vtwm ....

Help me out then. What value, to a window manager developer, is there in making it easier to create your own wayland compositor?

Fair enough, but it's not helpful either.

Nothing stopped the adopters from waiting until that existed before pushing Wayland into their software and breaking many people's workflow either, yet here we are

Do you mean default compositor?

As a FVWM daily driver, it's amazing to see it's has gone from "it's the niftier-than-twm baseline that's installed by default in your 2.0-kernel Slackware or RedHat distro, but you'll probably install something trendy like AfterStep/WindowMaker/Enlightenment" to "It Has Powers That Cannot Be Recreated In The New Magic."

For me, the winning feature is FvwmButtons. Long before we had system trays and notification busses, if you wanted to put a media player, a clock, some stat counters, or a full-blown xterm, in a little desktop dock, you just captured a regular window. You didn't have to invent an entirely new category of "software designed to live as an icon inside someone else's ecosystem." I'm not aware of any compositor that offers anything like it-- it seems like the best we get now are ugly bars with a limited vocabulary of "we can integrate over some signaling bus with these three specific programmes and that's it".

I'll also lament the loss of bold, opinionated design. "Modern" compositors are either minimal to the point of nothingness, or insipid and generic. They don't look like the awesome UIs you'd see in old hacker movies, or the classic systems that were backed by 500 page HCI standards guides, they just look like the sort of UI you'd use in an textbook where you wanted to imply a GUI without anything specifically branded.

> I'll also lament the loss of bold, opinionated design. "Modern" compositors are either minimal to the point of nothingness, or insipid and generic. They don't look like the awesome UIs you'd see in old hacker movies, or the classic systems that were backed by 500 page HCI standards guides, they just look like the sort of UI you'd use in an textbook where you wanted to imply a GUI without anything specifically branded.

On Mastodon I follow a bot that posts screenshots of old Mac Kaleidoscope schemes and the creativity on display both leaves me in awe and makes me sad that no modern windowing system can hold a candle to it. With Kaleidoscope there were no rules. You could do a hacker OS[0], or game UI[1], or titlebars on the left[2] or underneath[3], or non-rectangular and chrome[4], or made of denim[5], and those are just a few of the thousands of wildly different themes[6].

Nothing on modern Linux comes close. Even if you seek out third party themes all you find are dozens of minor permutations on popular flat themes like Material and Nord. It's so dull.

[0]: https://macthemes.garden/themes/24c0f39f11eb-net-ghost/ [1]: https://macthemes.garden/themes/8f7b145a33f3-Millenniac [2]: https://macthemes.garden/themes/95203ae3bfe0-my-sidewyas-os/ [3]: https://macthemes.garden/themes/6da656d65263-modulus/ [4]: https://macthemes.garden/themes/7e008718df3c-dt-chromxium-tw... [5]: https://macthemes.garden/themes/0305d1075f5e-dt-denim/ [6]: https://macthemes.garden/

> When Wayland replacing X, lots of cool window managers and mini applications will be gone.

There's hope due to the recent x11 fork, xlibre. They intend to keep x11 support ongoing

I think they meant like a reference implementation, wlroots came a decade into Wayland

This reminds me of LiteStep, which was super cool in the 00s, but to be fair, it got boring fast.

Alas, wasn't something like this also possible with enlightenment on Linux? I think there's experimental Wayland support.

I mean they should have made Weston production grade and extensible so that others can integrate with it instead of remaking their own compositors.

aka Wayland will never end up replacing X

But we also have wonderful new ones like niri

Application toolkits will eventually drop support for X11. GTK will remove it in GTK5. Not sure what Qt's plans are, but I'd have to think X11 support is long for this world there as well.