Mir vs Wayland: Choice is Good

So a few days ago, Phoronix published a performance comparison between XMir and X.org. Not surprisingly, XMir was slower in most of the tests. I say not surprising, because at this point its beta code (maybe even Alpha?) and is relatively new. Also, because it requires a compatibility layer for existing X applications.

When they asked the Mir team, Kevin Gunn said,

“at glance nothing looks completely out of place….except maybe the one nexuiz run, seems odd to be almost double. I suppose we might need to dig into those with >10% delta & determine the limiting factor.”

Or perhaps in emotes: *shrug*.

As a developer, I can appreciate that. I’m sure they are not focused on performance at this point and they have months to go.

Of course, a XMir review requires a follow up review of Wayland versus X.org. And so they did.

What I find interesting is that despite Wayland not being ready to run all the tests, including any 3D tests, the author concludes:

“I’m not an expert but I believe this may serve as an evidence that Wayland is architecturally superior than Mir.”

Huh? What? I have tremendous respect for Phoronix and their articles are great. They’ve forgotten more about this stuff than I’ll likely ever know. But that was a reach I think. It’s a bit … anxious ;)

UPDATE: Thanks to Assa for pointing out that Phoronix was simply quoting the tester and this was not their opinion.

Anyhow, I will be less than proud of the Linux community over the next 6 months as the FUD flies and the typical Ubuntu haters come crawling out of the wood work. Let me be clear: I am not saying Phoronix or this article falls into that category. It just made me sit back and realize that there are going to be people who are salty about the Wayland/Mir decision Ubuntu made. I personally like to see innovation and applaud anybody doing anything new and anything that suits their needs. It’s what Open Source is all about. Wayland or Mir, Unity or MATE, it’s all good.

But let’s not jump to conclusions in an effort to declare the next Ubuntu endeavor a failure. Let’s let it breath and see what goodness they will have for the open source community. And if you still don’t like it, great, there are lots of other great options.

6 thoughts on “Mir vs Wayland: Choice is Good

  1. I think it is a citate from Chang Liu, the guy who apparently did the test.

  2. I agree: stating that Wayland at this point is architecturally superior to Mir is a pretty unrealistic reach, but I think the more interesting thing is that Canonical decided to build Mir from the ground up instead of contribute to Wayland. Wayland may in fact be architecturally superior to Mir for the general user, simply by virtue of its development environment, but for Unity? It’s guaranteed that Mir will serve Unity better than any other display server, since the two are developed by the same team. Canonical made a choice to cater to its own services rather than open source at large. It’s going to benefit Ubuntu, sure, but Canonical’s resources could have turbo-charged Wayland’s development. With Mir, we will have a display server that has been designed around a user interface, rather than a user interface that has been designed around a display server. Kind of silly in my opinion.

    Then again, on second thought, I’m not sure I would want Shuttleworth and the rest of the Ubuntu crew to mess with Wayland. I appreciate his purpose, ultimately, but some of his ideas regarding software design and in particular “cadence” (you don’t need the latest version) make me a bit uneasy.

  3. I would be inclined to think that dropping dependence on Gnome would be a longterm goal for Ubuntu, but they will have a lot of ground to cover. The biggest obstacle to toppling X is the dependence between GL on Linux and Xorg. The lowest level binding to GL is GLX, which, predictably, pulls in Xorg dependencies. If canonical were to develop a display-server independent protocol for performing GL in a full OpenGL environment (i.e. not GLES), then that would be a huge service to the Linux, Mir, and Wayland communities. I’m not exactly sure how such a thing could be done, though, with the current GL architecture. GL operates in such a way that it uses X to pass display information to the monitor. We need an architecture by which the display server retrieves information from GL in such a way that it can be done agnostically, such that no program has to depend on the other.

Leave a Reply