Hacker News Reader

Comment by p2t2p | Hacker News Reader

In which world replacing removing Wayland in favour of X is considered KISS? We seem to have very different understanding of KISS with those folks.

View on HN

Wayland by itself is useless, you need other stuff to get some useful functionality, and the combination of Wayland+other stuff not only give you just a fraction of the functionality you get out of the box with Xorg but they also tend to be bloated silos (in that since Wayland itself doesn't do much it relies on extensions to do actual work but different Wayland implementations use different extensions - in theory X is the same, but in practice everyone has settled on Xorg and even without extensions, X provides enough functionality to allow creating useful applications).

Also for all practical purposes with Wayland you also need XWayland, which means that you get X anyway (especially considering that a lot of simple stuff will rely on X functionality). So might as well stick with X itself and get rid of Wayland as it doesn't provide anything of real value.

X is widly used since 30 years. Its configuration (and security model) is well documented (xorg.conf). It is as simple as launching the X server using a simple command line and being able to connect any client (even remote) thanks to a single environment variable (DISPLAY).

How do you launch a wayland server and how do you forward "display" (ssh -X or -Y) to launch a remote application ?

agree, even though some features of X are a bit crappy built, it's very well known, documented and used for so long basically any issue you run into someone else will have solved already somehow. perhaps one day wayland will be nice ,but not yet imho.

As for display forwarding via SSH, you may be interested in this summer's GSoC project Waypipe[1].

[1] https://news.ycombinator.com/item?id=20848334

Thank you very much. The work seems very complex. At the moment, I am at work on a windows PC. I have exceed and xming started and I have a toolbar with around 70 putty shortcuts with X11 forward. I am using them all the day long. At my previous job (using HP servers), it was a similar setup (with only 5 shortcuts).

Well, by that logic we should probably use COBOL everywhere.

I am lacking a proper analogy here, but it isn't because something is old, and most people are accustomed to its quirks, that it is simple... Maybe Apache, PHP, or even IPv6 would be a better comparison? IPv6 is likely a very good analogy here.

Now, Xorg's "security" model is well understood, just like its security issues. xorg.conf itself is a configuration nightmare. You might be able to launch the X server with a simple command line, but that's just hiding complexity behind a command. I can do the same with sway.

Wayland is quite modular itself. You can probably use namespaces to restrict one app's access to the clipboard, if that's not a feature the compositor supports, for instance.

For network "transparency" under wayland, which is an argument X proponents often cling to (it has its merits, but was counterproductive for most apps), it is a matter of serializing the protocol+image data (that's normally sent trough shared memory), and deserializing it at the other end. That's basically what waypipe [0] is doing, trough Unix sockets. So Wayland can be used remotely over anything that supports Unix sockets, thanks to waypipe, that plays the role of a compositor proxy.

A lot more things are possible due to Wayland being properly specified. The first and most important being multiple, interoperable implementations. Just like the TCP or HTTP stack. That is a Good Thing™ in my opinion. Zero-copy is also simpler to achieve, as another example.

I am sorry if I sound condescending about network transparency for instance. But that was pointed out for so long, and it is even written in Wayland's FAQ. The only reason it was that slow to materialize is Wayland's slow adoption. Which might be in part due to people complaining about stuff like network transparency without understanding it, and making themselves impervious to reasoning, while actually undermining efforts to bootstrap the ecosystem.

> How do you launch a wayland server

A compositor that displays stuff that applications send it? Well, you just execute it. Then that compositor decides to merge that stuff together, or pick some of it, and displays it where it wants: waypipe sends it to a socket, traditional wayland compositors can render on a connected screen trough the DRI kernel interface, others such as sway can even pass the rendered data to an X server or wayland compositor.

> thanks to a single environment variable (DISPLAY)

Use WAYLAND_DISPLAY to specify a Unix socket, if supported[1]. I'll admit that I am a bit in a learning phase, so I don't yet understand everything about the various subtleties (I couldn't use that to start an app from one TTY on another).

So, this distro sounds interesting on paper, but I can't really justify this choice of their, and I am probably not going to use it unless their revise their position. Wayland really needs a bit of forced adoption, like IPv6, if we want to have nice things.

[0]: https://gitlab.freedesktop.org/mstoeckl/waypipe

[1]: https://wayland.freedesktop.org/docs/html/ch04.html#sect-Pro...

The big problem with Wayland is that it isn't actually an obvious improvement over X. X is awful and long in the tooth and well overdue for replacement, but the best I can say about Wayland is that it is just differently awful. It's been in development for what, over a decade now? And it still would be missing a lot of basic functionality if it weren't for people like Drew Devault stepping up and creating new standard protocols to cover for it, as well as a library to make dealing with all these protocols less of a nightmare.

I'm not sure which point you're trying to make. Of course every new technology needs someone to do the work. That doesn't take away from the technology's merit. If Drew is now doing all this work, then great! Now is the best time ever to be adopting Wayland.

(Speaking as someone who just switched to Sway two weeks ago.)

My point is that in setting out to replace X, they created something that doesn't actually do that and it took about ten years for third parties to do it for them and there still isn't consensus that Wayland is superior to X.

In comparison, Windows Vista introduced a perfectly functional compositor with brand new driver model at about the same time and I can't think of any significant functionality it lost in the process, while I can name many significant improvements.

> I can't think of any significant functionality it lost in the process

Hardware acceleration for 2D operations (GDI)

In what aspect Wayland makes more KISS?

You cannot completely remove X anyway, so using Wayland inevitably introduces much compatibility headache. Just sticking with X is the simples and KISS way to get GUI for now.

Wayland has a less powerful API, multiple frames of latency, and last time I checked, was locked to 60hz and the devs don't care

Reply on news.ycombinator.com