Told you so, Mir/Wayland are a dead end

Story: Benchmark: X.org VS MirTotal Replies: 14
Author Content
linux4567

Jul 18, 2013
12:24 AM EDT
Hah! Wasn't one of the selling points of Mir/Wayland supposed to be that it would be a lot faster due to trimming all the 'fat' that X allegedly carries?

So in reality X is still faster than Mir/Wayland and it's far more feature rich.

I knew it that all the talk about bloat was rubbish, X isn't bloated, or to be more precise the so called 'bloat' (code that's no longer needed) doesn't affect performance as in most situations it doesn't get used anyway.

I much rather stick with X due to it's features, first of all network transparency and now I know that I'm not even compromising on performance by sticking with X.

Just like Pulseaudio, Mir/Wayland seems like another 'solution' looking for a problem that doesn't exist.
rahulsundaram

Jul 18, 2013
12:35 AM EDT
Couple of major problems with the previous post

1) Wayland developers ARE Xorg developers. Xorg foundation considers Wayland as a successor to Xorg. Performance is hardly the main criteria for Wayland. Feel free to watch http://www.youtube.com/watch?v=RIctzAQOe44 to understand the reasons

2) The article doesn't compare Xorg performance and Wayland at all. Only Mir. So your conclusion is premature to say the least.
penguinist

Jul 18, 2013
8:56 AM EDT
For an interesting comparison of Wayland and X11 running on low end hardware, watch this demo:

http://www.collabora.com/videos/rpi-wayland-demo-720p.webm
linux4567

Jul 18, 2013
9:18 AM EDT
> Wayland developers ARE Xorg developers

I know that, so what's your point?

> Performance is hardly the main criteria for Wayland

Performance has always been mentioned as a major reason for Wayland, denying that now is beyond silly (and either way, users will have no interest switching to a slower system that on top of that lacks important features).

> The article doesn't compare Xorg performance and Wayland at all. Only Mir.

Given that Mir is a fork of Wayland it's reasonable to assume that perforamce will be similar.
CFWhitman

Jul 18, 2013
9:37 AM EDT
There are more problems:

3) To get 3D graphics for these benchmarks under Xorg, you use drivers that basically circumvent Xorg. That can work to get good performance, but it complicates matters for driver, Xorg, and Mesa programmers.

4) As a related point: basically, all the things that Xorg doesn't do or does very badly are instead done by window managers, drivers, and other software. A lot of stuff that X was written to do, programmers purposely avoid letting it do.

5) Even if Wayland has similar performance to Mir (and thus similar to Xorg now, if you can really call these benchmarks Xorg performance (remember that they're circumventing most of X)), Wayland is beta software and hasn't been optimized.

6) Wayland isn't intended to completely replace everything about Xorg. Parts of Xorg that are worth keeping can be kept.

The worst thing about replacing Xorg is all the software in existence that expects an X server. That is why we will probably need to keep an X server or a pseudo X server around for a while. However, in the long run it makes all the programmer's lives easier to write for something like Wayland rather than always trying to compensate for deficiencies of X. X has a network layer; that's great, but it's not worth keeping all of X around just for its network layer.
Fettoosh

Jul 18, 2013
11:05 AM EDT
Quoting:Given that Mir is a fork of Wayland it's reasonable to assume that perforamce will be similar.


@linux4567,

That isn't and can't be true, different needs, focus, and code. But, If it makes you happy and nothing is going to convince you otherwise, keep using Xorg. We others are moving on to better things.

EOS (End Of Story).

skelband

Jul 18, 2013
12:07 PM EDT
The aims of Wayland include speed, but it is one of many issues, the main of which seems to be a combination of compositor and appropriateness for modern thick-client use cases.

As CFWhitman says above, most of what XOrg does is not used by many desktop applications and it just gets in the way.
penguinist

Jul 18, 2013
12:31 PM EDT
The use case for "network transparency" seems to have largely disappeared over the decades.

These days, network security trumps convenience in importance. For me, 95% of my remote operations are through an ssh terminal connection, another 4% go over an "ssh -X" connection, and the remaining 1% run with vnc over ssh. I personally have no need for X itself to provide network support.
CFWhitman

Jul 18, 2013
2:28 PM EDT
After looking into it still further, I see that even the network layer of X is outdated. Wayland can/will get network capabilities that are more efficient than those of X. Even now, you're generally better off with VNC than using X over a network most of the time.
linux4567

Jul 18, 2013
3:12 PM EDT
> The use case for "network transparency" seems to have largely disappeared over the decades.

That is not true at all, a simple home user might have no need for it, but it's still in common use in companies that use Unix systems (not just Linux) internally

> another 4% go over an "ssh -X" connection

Do you realise that 'ssh -X' is using the network transparency feature of X? All it does is it tunnels the X server connection over ssh, but it will only work with X, not with pure Wayland or Mir that don't support the X network protocol.

That's why getting rid of fundamental features such as network transparency is so wrong, many people like you use it but don't even know they are using it.

> Wayland can/will get network capabilities that are more efficient than those of X

That's the first time I hear that, do you have any sources for that?

So far all I have read is that network transparency is beyond the remit of Wayland/Mir and people should use remote desktop solutions like VNC which are something very different, they are nowhere near as practical as simply forwarding the display of a single application to a remote PC.

Don't get me wrong, VNC is fine for it's intended purpose (to remotely access an entire desktop) but it's something very different than the network capabilities of X.
fnoss

Jul 18, 2013
3:22 PM EDT
@linux4567 and all

This benchmark is *not* xorg vs Mir, it's xorg vs *Xmir* see the quote below.

"Final Results:

As shown by the test, X.org beats Mir in all cases. As explained on Google+ by +Francesco Corazza, the version of Mir currently present in Ubuntu 13.10 is not “clean”. it’s called XMir because it offers a compatibility support for applications that still talk with the “old” X.org protocol. The operations then pass both Mir and X.org: this explains the slight drop in performance."

The article is very misleading up to that point.

Wayland will be faster then xorg as the protocal stack has far fewer hoops to jump through then xorg does.



linux4567

Jul 18, 2013
3:29 PM EDT
> This benchmark is *not* xorg vs Mir, it's xorg vs *Xmir* see the quote below.

That is irrelevant as pure Mir/Wayland won't exist for many years to come as otherwise older apps written for X (and any new ones that people who don't care about Mir/Wayland will still write for X) won't run on it. Mir/Wayland will have an X compatibility layer for many years to come.

Also Francesco's argument that this is the reason for the slowness is IMHO flawed as the X compatibility layer runs on top of Mir/Wayland (not underneath), so native Mir/Wayland apps don't make use of it therefore it shouldn't influence performance of native Mir/Wayland apps.

As far as I understand it the apps that were benchmarked (or rather the GUI libraries they depend on) were built natively for Mir/Wayland so they should not be affected by the X compatibility layer as they completely bypass it.
fnoss

Jul 18, 2013
3:53 PM EDT
>That is irrelevant as pure Mir/Wayland won't exist for many years to come as otherwise older apps written for X (and any new ones that people who don't care about Mir/Wayland will still write for X) won't run on it. Mir/Wayland will have an X compatibility layer for many years to come.

X applications on a native wayland compositor, is not tha same as a full x compositor running on wayland.

x applications running in their own xserver atop wayland are able to miss some crutial steps that a full desktop can not. don't ask me what steps as I don't know, but I do know this is true.

>As far as I understand it the apps that were benchmarked (or rather the GUI libraries they depend on) were built natively for Mir/Wayland so they should not be affected by the X compatibility layer.

they are still running on a compatability layer. so this makes absolutely no difference. Also, where does it say this? Is it your article?
CFWhitman

Jul 18, 2013
4:35 PM EDT
linux4567 wrote: That's the first time I hear that, do you have any sources for that?


Daniel Stone (a former X and current Wayland developer) talks about it toward the end of the video that rahulsundaram posted.
lietkynes

Jul 18, 2013
4:48 PM EDT
Classic case of someone counting their chickens... We can't get any information from this...Your jumping on any bit of doubious evidence to fulfill your profecy is cute; but irrelevant. If you truly want to miss the point of Wayland/Mir, you are welcome to. However, note how no one cares. We all want different things from Wayland based on our experience. More elemental is the fact that Developers want this on all camps and their reasons are not speed. I could list the very tangible benefits...but why be redundant?

PS: No one more is backtracking based on this article. In part because it doesn't show anything....but fundamentally because speed is a small added bonus.

Posting in this forum is limited to members of the group: [ForumMods, SITEADMINS, MEMBERS.]

Becoming a member of LXer is easy and free. Join Us!