Hacker News Reader

Comment by djsumdog | Hacker News Reader

Xorg doesn't really need a configuration file today; it's good at auto detecting hardware. But one of the things I really like about X is that you can have a config file and specify exactly which devices you want. In this way, you can connect two keyboards, two mice and two video cards to the same machine and have two totally different login screens. You can essentially have two people logged into the same machine as if they were using two different terminals (just do try to plug in a USB device if both users have automounters running).

Even on my current single user machine, I still run multiple X servers (one for my work related stuff and one for non-work. I work remotely so that way, I can just switch X servers at the end of the business day).

Can Wayland support multiple terminals on a single machine in the same way? Can you have two Wayland instances running like you can with X servers?

View on HN

Wayland (or more "correctly" the support plumbing that will surround it, but developed by the same team) will likely support all that, in time.

Just that rather than manually configure it, it will be automatically (they hope) configured by combo of systemd, polkit and your DE of "choice".

So once again freedesktop.org pushes out the new software with the promise that some day it will meet the capabilities of the current tools, honest, we pinky swear. They did it with systemd and they'll do it again with Wayland.

Yeah, ignoring those seldom-used edge cases (multiple mice, keyboards, N people using the same computer at once, etc.) all the while solving several real issues and improving common uses cases in the process...

Damn those freedesktop.org people!

Yes, we know the Gnome team (and by extension, most of Freedesktop) wants to reimplement macOS on Linux. But trust me, Miguel isn't coming back, the best course would be just going to macOS instead (or hell, Windows, he's a happy camper now, isn't he?) or do your work downstream.

The rest of us just want our own systems to run as well or better for our own usecases as they currently do. Wayland is a great opportunity to rearchitect the Linux display system rather than organically grow it as X11 has over decades. But if you don't take into consideration those of us who use our Linux workstations as more than a Facebook machine with SSH preinstalled; you're just going to recreate the problem.

People will start disregarding freedesktop, Wayland and so on and either stay in X or using a Wayland compositor that couldn't care less for standards and short-circuit every single convention and security protocol you have designed as long as it lets them get shit done, because at the end of the day, as beautiful as we like our code to be and our software to be architected; people want their systems to work for them.

> The rest of us just want our own systems to run as well or better for our own usecases as they currently do.

Well, speak for yourself. Which seems to be the loudest voices in these debates, people speaking for themselves.

Wayland solves real, systemic issues in windowing, most notably around security, that need to be solved. It also allows outsourcing of many of X's most unwieldy features to the clients (IPC, rendering, and networking). In many ways, Wayland is closer to the Unix philosophy of one program/one task than X while also being safer to use.

And, it's not like you need to drop X, just like you don't need to drop init. The beauty is that you can continue to use and maintain whichever you feel are the best tools.

> People will start disregarding freedesktop, Wayland and so on and either stay in X or using a Wayland compositor that couldn't care less for standards...

I feel like the same things were said about systemd a few years ago and here we are, every major distribution uses it out of the box. And every windowing system and graphical toolkit supports Wayland, long before Wayland was a default. Though I'm not an expert in graphical compositing and rendering, I can't imagine this is by accident. Numerous projects must feel the technical advantages of Wayland outweigh the disadvantages of change.

> I feel like the same things were said about systemd a few years ago and here we are, every major distribution uses it out of the box.

Say hello to RH fueled code churn. First it was powerkit, then it was powerd, and now most of it is found inside systemd-logind. Want to power down the computer by clicking a button in the DE, you need systemd-logind, powerd as a wrapper (for now), polkit, and dbus to glue it all together.

That or you can forgo building your default distro install around any of the major DEs.

The CADT that JWZ railed against a decade ago have gone supersonic, and heading towards superluminal.

I really fear for the day Torvalds hands off the maintenance of the kernel, as he is likely to hand it to someone that is already very accepting of this churn.

Edit: BTW, it seems the BSDs still rely on hald. Maybe i should bite the "apple" and move on over.

>Wayland solves real, systemic issues in windowing, most notably around security, that need to be solved. It also allows outsourcing of many of X's most unwieldy features to the clients (IPC, rendering, and networking). In many ways, Wayland is closer to the Unix philosophy of one program/one task than X while also being safer to use.

That's exactly my sentiment and I brought it up in the comment. The problem is designing these protocols without regard for how the thing you're replacing is being used. That's only going to result on a pile of hacks. It's just baffling how short-sighted some of the Wayland devs are when the project was born precisely because of these hacks accumulating in X.

> And, it's not like you need to drop X, just like you don't need to drop init. The beauty is that you can continue to use and maintain whichever you feel are the best tools.

Oh yes, because obviously everyone wants to maintain 10 different code paths so it's not like Red Hat is pushing their software with their influence over the ecosystem oh no no. But enjoy maintaining those shims, we sure as hell don't care about standards! Gotta move fast, break things and compatibility is for losers.

>I feel like the same things were said about systemd a few years ago and here we are, every major distribution uses it out of the box.

And it's just been dandy isn't it? We still don't have stable interfaces between its parts (yet they still have the gall to say that systemd isn't monolithic), good effin luck debugging it when it shits the bed... list is long, but pick your poison: http://without-systemd.org/wiki/index.php/Arguments_against_...

Negative points will be awarded on "works for me".

> And every windowing system and graphical toolkit supports Wayland, long before Wayland was a default. Though I'm not an expert in graphical compositing and rendering, I can't imagine this is by accident. Numerous projects must feel the technical advantages of Wayland outweigh the disadvantages of change.

The premise is great, what's not to love about it. The problem is execution. Because of Wayland's model, a lot of applications relying on X's nature are left out in the dirt. Standards are the way to solve this, and some work is being done about this, but the comment I replied on by coldtea simply shat on all this with a message akin to "Don't stop the momentum, you're a minority FUCK YOU we don't care about you get on board or get ran over". This is what I was criticizing.

We need standardization. Ideally we have a single solution that improves on the status quo and we call it a day; but where there's disagreement let's find our middle ground and standardize it. If you don't need 4 mice and 4 keyboards working simultaneously great! But what you find useless and nonsensical can be the greatest fun ever in a 4-player co-op session of Serious Sam 3 for groups of friends around the world; or just as well prove to be one of the most productive formats for groups of people working together in a local application.

Case in point: the core Wayland protocol desktop applications use for creating and handling windows doesn't support basic functionality like minimizing them, because Gnome doesn't think people should be able to minimize windows. So there's a complete duplicate of the Wayland window creation protocol as an extension to add support for features like this. Actually, effectively several complete duplicates since they keep changing the protocol in ways that aren't backwards compatible whenever they need to add new features.

> Oh yes, because obviously everyone wants to maintain 10 different code paths

Can you really blame X developers for not wanting to maintain something painful?

It's fair for them to focus on things that will please the majority of users. Not everyone are happy with using 40-year-old technology for their daily needs. But this is open-source; contribute if you like it.

> We need standardization. Ideally we have a single solution

X has its own barely readable standard, and Wayland is a separate standard protocol which is unavoidably incompatible. The core X protocol makes assumptions that are no longer valid, and its maintainers have tried patching around it for dozens of years without breaking the standard, but many issues were simply unsolvable.

I think you misunderstood me. The people I had in mind were those maintaining code that writes its own windows (toolkits, libraries like SDL2, plenty of games rolling their own and applications like Steam).

And I mentioned it plenty of times: But I like the idea behind Wayland, and think it's a great thing that the display system is finally getting rearchitected and rewritten.

My problem isn't "I no longer have X", my problem was the attitude of "Wayland doesn't have to care about non-average users when replacing X" which I was previously responding to.

Wouldn't supporting simultaneously the X protocol and the Wayland protocol be very hard, if not impossible? It's already hard enough to replace the "basics" of X.

Not impossible, just ugly as hell. The first two examples I mentioned (toolkits like Qt or Gtk and the SDL2 libraries) actually do it.

> Yes, we know the Gnome team (and by extension, most of Freedesktop) wants to reimplement macOS on Linux.

Ugh, there actually are people who enjoy the GNOME desktop, so maybe don't present your personal opinion as a collective sentiment. As an owner of a MBP, I can tell you that I prefer GNOME in speed, workflow, look, customizability - you name it. In what way exactly is GNOME like macOS in your opinion?

What happened to Miguel? Did he get deported?

He works for Microsoft, on Xamarin. And seems happy doing it. Hence the statement that he won't be working on GNOME again.

improving common uses cases in the process

The whole reason people switch to Linux in the first place is for uncommon use cases.

>The whole reason people switch to Linux in the first place is for uncommon use cases.

On the server, that's not true. People use Linux for the most commonplace use cases (e.g. the LAMP stack) as well.

On the desktop, that's what needs to change, and has been promised to change since 1999 ("next year will be the year of the Linux desktop" which has been endlessly repeated was not about some outliers using it for "uncommon" use cases, but about mass adoption).

But the whole critique is wrong too. When I said about "improving common uses cases" to the detriment of "uncommon uses cases" the former wasn't referring to features different people use. The people who use those "uncommon use cases" ALSO suffered from the crappy implementations and shortcomings of the common use cases like everybody else.

On the server Linux was basically carried in by LAMP, as it allowed people to spin up websites at only the cost of the hardware (or rent on server space). Never mind that this as the dot-com bust, and thus there was a whole lot of cheap capacity to be had (kinda reminds one of the microcomputer getting a boost from excess capacity of CPUs thanks to the moon race).

On the desktop Linux have a twofold problem. No prominent presence in stores (observe that Android, being a Linux derivative, have a massive presence, and the rise of ChromeOS) and a lack of commitment to API stability from the major lib devs (again contrast Android, where you have versioned APIs going back to the early days).

Really? So nobody's using Linux as a traditional desktop because they like it better than what Microsoft and Apple have to offer, do they?

Well, I beg to differ.

We live in a world of touchscreen laptops, and you consider 'multiple mice' to be an edge case?

Interesting.

I'm not sure I following the reasoning process (or lack thereof) used to make this association.

First, just because something is common (multitouch laptops), doesn't mean something similar is equally common (multiple mice).

Second, those multiple mice weren't used for a person controlling multiple pointers at the same time, but for multiple people using the same computer in different window sessions. So not like multi-touch at all.

Third, even if multiple mice used at the same time by the same person was what we were talking about, it doesn't follow at all that it would be as convenient and approachable as multi-touch to have it be something more of an edge case.

The edge case isn't multiple mice, it's multiple mice controlling multiple pointers in multiple sessions.

> So once again freedesktop.org pushes out the new software with the promise that some day it will meet the capabilities of the current tools

You can use Xorg as you always did, Wayland is fully optional.

> They did it with systemd

What part of hacky, inconsistent, non-dynamic and hard to maintain InitV shell scrips do you actually miss in systemd? Or do you just hate it because it's cool, like it was with PulseAudio?

The alternative to systemd was not the old init-v, but upstart.

Not just: http://without-systemd.org/wiki/index.php/Init

Upstart was being pushed by Ubuntu. Systemd was being pushed by RedHat.

The others where just files in some server, used by almost nobody.

Ohh, and in some cases even the files have been lost: http://git.einit.org/

> You can use Xorg as you always did, Wayland is fully optional.

For now...

Wayland allows for sane graphical rendering pipeline. It's missing in X. A tiny raspberry pi becomes silk smooth by switching to wayland.

I don't dismiss your points at all, just showing wayland is not a subset of X.

Wayland has been around for 4 years, When X was 4 years old, it ran on hardware with 5mhz processors and a few MB RAM.

> Just that rather than manually configure it, it will be automatically

That's the point. It'already automatic.

I haven't even mistakenly opened xorg.conf in at least five years.

And quite franky, I see no need for another thing like Wayland to add entropy to the GNU/Linux desktop.

> by combo of systemd, polkit and your DE of "choice"

eh...

First of all you have systemd-logind that tracks "sessions" (using a different definition to the one found inside the kernel, natch) and "seats", that then polkit talks to via dbus (yay...), and then the DE talks to polkit (again via dbus).

These then try to tie a video output, and at least one input device (mouse, keyboard, whatsnot) to a specific seat (as defined by systemd-logind) so that only associated (systemd-logind defined) sessions can interact with them.

Note that all this happens above the kernel, that (as best i can tell) do not have a concept of a seat, and sessions are tracked on a tty basis.

And ttys used to be physical devices connected to the computer via a serial connection (and still can be even on Linux, nice for getting those early kernel errors onto something that do not go poof the moment the GPU driver code do a belly flop).

Never mind that X was designed to work as a graphical tty of sorts. Thus rather than try to stuff raw bitmaps down the wire, the client code running on the computer would send draw command primitives (line from X,X to Y,Y, square with top left corner at A,A and bottom right at B,B, that sort of stuff) that the X server sitting on a desk somewhere would turn into a GUI.

I think Xscreensaver still use that for drawing the "screen lock" password prompt and launch popup.

>And quite franky, I see no need for another thing like Wayland to add entropy to the GNU/Linux desktop.

I'm sure the merits can be discussed, but security is a need that Wayland addresses.

I know you are making an earnest comment but the idea of having one computer support multiple users each on their own terminal is really really well explored territory. So I chuckled a bit at how even old things can be new again.

It does speak to having way more compute resources than you can consume at a given time.

Its old yes, but the Freedesktop boys are doing it in a "new" way. This by using sofware to bundled a random set of devices into a single seat, and assigning said seat to a user on login based on the input device used. The old way was to hang dedicated hardware off serial ports and figure out the "seat" (if such a concept needed to be defined at all back then) that way.

Exactly right, the concept is timeless and the implementation choices change with the times.

A similar moment of humor occurred when AT&T's U-verse product offering was bragging about sending your video signal to other TVs in your house wirelessly.

So the next engineering question is "How does this make things we want to do, or have done, better or different?" I could imagine you could set up a very interesting pair programming setup that way, especially if you were projecting a common "third" monitor on the wall or something. That common monitor could show current status for both programmers like building/testing, source code state, etc.

With Wayland I can finally dock & undock my thinkpad without having to restart my laptop.

I can do the same with Xorg (also thinkpad), what kind of problems do you have in Xorg with docking?

Screen was freezing in Fedora. In Ubuntu it was working but took ~25 seconds for screen to come on. constant problems across several OS releases & versions.

What configuration? Works fine for me with a Skylake ThinkPad on Fedora 24.

Reply on news.ycombinator.com