Canonical writing their own X/Wayland replacement | Hacker News Reader

Canonical writing their own X/Wayland replacement

lists.ubuntu.com

68 pointsshadeslayer38 comments

Ha, I bet this one will crash and burn over Fiji too.

More information here: https://wiki.ubuntu.com/MirSpec

They don't like X for all the normal reasons, they think Wayland might be like X but haven't investigated (I think you can replace the input handling in the "compositor" plugin; maybe I'm wrong).

They don't specify how they'll handle hard problems, such as:

- Client vblank sync (as well as compositor vblank sync). Compiz, unity, etc, don't do this now; clients run unsynced and tear.

- Switching unoccluded windows to render directly to the framebuffer instead of going through the compositor. Mandatory for good performance on something like a Nexus 10 (which has a 16MB framebuffer, so nearly 1GB/s bandwidth to deliver glClear at 60fps with no extra copies...).

- Handling multiple clients that want to read gestures. They actually say in their "INPUT" section that they will handle "shell" gestures "in the server", which sounds like they're going to bake their UI into their display server.

They want it to be compatible with all of the old apps too, which Wayland already does. I'm trying not to be judgemental, but it sounds like they want to write something like Wayland but without designing it first. Good luck with re-entry, guys.

Bigger burn than reentry.

You're right though, this feels a bit like not developed here syndrome. X is horrible to hack on top, and has had a good run but Wayland could be a fantastic base for them, especially given Qt has spent time working on it.

Is it possible that once toolkits like qt and gtk have been ported to simpler/more direct APIs like wayland (w.r.t X11), porting them to other APIs like this mir would be easier than starting from scratch?

Possibly, but I've followed some of the Qt Wayland development and it looks like they've hit problems on a fairly regular basis. When you're building GUI toolkits you're also including a certain amount of understanding of the weird corner cases from the windowing system, so you're ending up porting a large chunk of code and then working out what's broken.

Although, Qt might have it a bit easier as they've switched to drawing their GUI components with OpenGL.

They were the major ones touting Wayland for a while, it seems unlikely that they "haven't investigated".

Other than, e.g., Red Hat[0]?

[0]: https://en.wikipedia.org/wiki/Wayland_(display_server_protoc...

Open source is wonderful. Good luck to Ubuntu, though I love X. As an end user of X, fluxbox, and X-forwarding over ssh, X just works. I'm not looking forward to compatibility troubles when the new fleet of X replacements arrives.

Debian, please continue to maintain old stalwarts as Ubuntu and other distros flirt with the new.

What happens to the X windows forwarded over ssh when the connection drops? Can you reattach afterwards?

No, I think the app gets killed.

Depends on the app. Emacs can survive and re-attach later. A single emacs session can be connected to multiple X (and tty) sessions at a time. I use this all day, my emacs session runs for weeks or months between restarts and I just connect/disconnect to X servers or ttys as needed.

If you use something like Xpra, sure: http://xpra.org/

I thought we had moved beyond these gripes about Wayland. Wayland already supports a far more performant remote networking ability and XWayland gives you all the compatibility you need.

You may want to go to https://wiki.ubuntu.com/MirSpec if you want to read about their motivations. I don't enough about Wayland to say if their argumentation is justified or not.

Somehow, I'm less than surprised. I'm waiting for them to decide to roll their own browser, email client, and office suite...

Don't forget about their own programming language. :)

Rubunthy on Mails?

[edit: Oops, I've jumped all the way to their own web development framework. At least this way they can also build their own DSL's.]

> their own browser, email client, and office suite...

Iceweasel, Icedove, LibreOffice? They seem to be working on it ;-).

Iceweasel is Debian's, not Ubuntu's; it is a patched version of Firefox. It is not called Firefox, because the license of Firefox doesn't allow you to patch it and call it Firefox. Icedove, likewise, but Thunderbird instead of Firefox. LibreOffice was OpenOffice which was StarOffice, which was developed as a commercial project, bought (along with its developers) by Sun, and released as FLOSS in 2000. The project has been through renamings and possibly reorganizations since, and I can't say whether Ubuntu has had any involvement, but to say it's an attempt by Ubuntu to develop their own office suite is grossly disconnected from reality.

It works for GNOME and KDE. ;)

KDE started ours way before the others came along (or in the case of the browser, shortly before). :P

They ought to fix all the pretty basic bugs and commit to actually supporting their customers before they bite this off.

LaunchPad is a testament to the shit state Ubuntu is in. All the windows are broken so we'll knock half the building down and rebuild it. Doesn't work.

...really?

X is an abomination. But why not Wayland? Anyone care to present Canonical's argument, with concrete examples of what they find so offensive about it?

https://wiki.ubuntu.com/MirSpec#Why_Not_Wayland_.2BAC8_Westo...

If Wayland's input handling is not good enough, why didn't they raise that on the mailing list any time in the last few years?

What does "The shell integration parts of the protocol are considered privileged from our perspective" even mean?

I think it's just as interesting (if not more so) that they are moving away from Gnome to Qt!

That bit is nice to see, but I'm pretty suspicious of it. They're going to have to write _a lot_ of software to fill the gaps, now, if they mean to have Qt everywhere. GNOME provides a lot of functionality beyond the UI toolkit and the shell, including myriads of small applications we take for granted. I'm curious where Canonical is going to find the engineering resources to handle all of this.

> including myriads of small applications we take for granted

As someone who runs Arch with KDE and an intentionally different Gnome theme just so I can notice the apps I use running Gnome, there aren't too many. In Ubuntu-space, definitely, since their entire ecosystem was built on gnome, but in general you can easily get a gnome-less Linux.

Gnome-less debian user right here!

Another item of note in this post is that they are moving Unity to Qt.

Did they give any rationale why they can't use Wayland?

Yes. Canonical clearly lays out the project motivations, including not choosing to use Wayland, on the Mir wiki page[1].

[1] https://wiki.ubuntu.com/MirSpec#Why_Not_Wayland_.2BAC8_Westo...

What worries me is fragmentation of drivers. It can be a very serious problem.

They'll probably use the same KMS and EGL drivers that Wayland and others use.

If that's the case - it would be good. But I'm not sure about it yet.

Yes, this is covered in the wiki page linked to in the email:

  https://wiki.ubuntu.com/MirSpec

from Wayland creator Kristian Høgsberg: " things they claim wayland/weston input can't be extended to support: "... adjusting and extending X's input model is difficult and supporting features like input event batching and compression, motion event prediction together with associated power-saving strategies or flexible synchronization schemes for aligning input event delivery and rendering operations is (too) complex." is already implemented and working in weston today..." https://plus.google.com/100409717163242445476/posts

(from other thread on Mir)

Saw someone post this log from #wayland on reddit. An cursory reading seems to indicate ubuntu folks don't fully understand wayland, and at the very least didn't even bother discussing their perceived gripes with the wayland team.

edit: forgot the link! http://pastebin.com/KjRm3be1

I'm surprised this hasn't been mentioned yet: Ubuntu was able to cobble together Ubuntu Touch running on Surface Flinger as a compositor.

There are limitations to that, including lack of really working support for multiple monitors.

Mir may be designed to fill in the deficiencies of Surface Flinger for use as a compositor for Ubuntu. That might make project scope controllable enough to think it could succeed.

Reply on news.ycombinator.com