Mark Shuttleworth: Mir By Default In Ubuntu 16.04 LTS

[profoundWHALE]

That was a terrible comparison; don't try to compare something like Wayland, Xorg, or Mir to the freaking Linux kernel ever again.

[sarmad]

But Fedora was never delayed because they decided that "hey, the Linux kernel just doesn't suit our needs any more somehow, so we're starting our own inhouse project, Finux, which is basically the same as Linux, does all the same things, but the difference is we get to change the APIs however we want and we will never stick to any protocol or worry about compatibility with other software".

QFT. Fedora was delayed because they didn't have enough time to do the necessary, not because their resources are consumed in doing unnecessary work like Ubuntu. If Canonical saved those resources and just used Wayland (or even just stick to X.org) they probably wouldn't have needed to delay their convergence plan.

[jumu]

QFT. Fedora was delayed because they didn't have enough time to do the necessary, not because their resources are consumed in doing unnecessary work like Ubuntu. If Canonical saved those resources and just used Wayland (or even just stick to X.org) they probably wouldn't have needed to delay their convergence plan.

I think the convergance time would be 100% the same with wayland. I am sure that not even 1% of their resources go into "protocol" definition.

- the new compositor they had to create anyway for the whole convergence transition
- unity 8 had to be a whole rebuild anyway for the convergence
- all the helper components have to be created anyway to achieve their vision
- the SDK has to be build anyway for the convergence
- actually not using could have accelerated the unity 8 development and DM work because as it goes and as needs come up they can just quickly design it to what they exactly need.

So please dont believe that the lack of wayland delays the convergence vision.

[cynical]

No, because there's nothing else to jump to.



But Fedora was never delayed because they decided that "hey, the Linux kernel just doesn't suit our needs any more somehow, so we're starting our own inhouse project, Finux, which is basically the same as Linux, does all the same things, but the difference is we get to change the APIs however we want and we will never stick to any protocol or worry about compatibility with other software".

Yeah way to completely miss the point of my post in order to bash Canonical yet again. Delays happen, predictions aren't accurate. That doesn't mean what you are working on is doomed.

[cynical]

QFT. Fedora was delayed because they didn't have enough time to do the necessary, not because their resources are consumed in doing unnecessary work like Ubuntu.

Yeah that's not subjective at all.

[dee.]

Yeah way to completely miss the point of my post in order to bash Canonical yet again. Delays happen, predictions aren't accurate. That doesn't mean what you are working on is doomed.

No. There are other reasons for why Mir is doomed. Like the fact that Mir requires its own graphics stack, which requires maintaining Mir-specific patches, and that Mir won't be compatible with Wayland clients, and then there's all the toolkits, requiring Canonical to patch in Mir support to toolkits, which requires maintaining even more Mir-specific patches...

That's a lot of extra work that wouldn't be necessary if they just used Wayland.

See, the real benefit with Wayland is that it's a collaborative effort. All the other distros and DE's are behind it, contributing code and/or supporting it, porting toolkits and applications, making necessary changes to the graphics stack... That's one of the main benefits of open source, it's the reason why the Linux kernel has gotten so popular - the most used kernel in the world and the largest software project in the history of mankind. It's because the open source model lets even competing companies collaborate on something, sharing both the workload and the benefits. It saves resources for everyone.

Canonical is doing the exact opposite, they're taking bits and pieces of other people's work, cobbling them together and adjusting them to make them incompatible with others. Mir wouldn't be even possible without all the groundwork done by Wayland developers, and Canonical is now taking advantage of that work in order to create their own private in-house solution, but they're missing out on the larger advantage of the collaborative development model.

[Rich Oliver]

Seems to me that Canonical doesn't really understand open source. They're constantly missing out on the main benefits of open source, they think it's just a convenient way to get code developed for their inhouse solutions. In short, they're trying too much to be like Apple, but without having all the money, massive resources for r&d, their own hardware line, etc... And even Apple is fading already, even with their huge resources - so why does Canonical, a small company that can't even afford to release a phone without a crowdsourcing campaign, think they can succeed with the same strategy? It's just plain foolishness with possibly streaks of megalomania.

Well I've been predicting for a while that the Self-Appointed Benevolent Dictator for Life would prove to be the Mussolini of Open Source.

[bregma]

Canonical is doing the exact opposite, they're taking bits and pieces of other people's work, cobbling them together and adjusting them to make them incompatible with others. Mir wouldn't be even possible without all the groundwork done by Wayland developers, and Canonical is now taking advantage of that work in order to create their own private in-house solution, but they're missing out on the larger advantage of the collaborative development model.

Well, given that Mir is free software and developed in the open, you should be able to point to those parts of other people's work you are accusing Canonical of taking and and making private. Please post links directly into the public archives.

You seem to be missing the main point of free software, which is freedom. Not freedom from choice, I didn't see that listed in the Free Software Manifesto, but freedom to use, study, modify, or replace. Being a slave to a single alternative is a very Orwellian notion of freedom indeed. We are all better served by having more choice, more alternatives. If a Wayland implementation works well going in to the future, that's great. If Mir works well going into the future, that's great. We all win. If only one alternative controlled by the interests of a couple of entrenched billion-dollar corporations is all we should have, I fail to see long-term benefit.

[Gusar]

Well, given that Mir is free software and developed in the open, you should be able to point to those parts of other people's work you are accusing Canonical of taking and and making private

XMir is basically a fork of XWayland. The much touted advantage of Mir, being able to use Android drivers, was made possible using a library that was developed for Wayland (libhybris). Then there's mesa EGL, mesa gbm, possibly other parts of mesa, and even KMS itself - they're not specific to Wayland, but they made Wayland possible, and likewise there couldn't be Mir without them.

I'm too lazy right now to give you specific links, so feel free to ignore the above. It is correct though, other not so lazy forum participants will confirm it and maybe even add more examples.

We all win.

Toolkit developers need to write support for two display servers, including the QA and maintenance burden that additional code brings with it. Driver developers need to support two display servers - while both Wayland and Mir are based on EGL, I very much doubt drivers will be free of display server specific code. Distro maintainers need to decide whether they'll ship both display servers, and if they do, this again increases QA and maintenance burden. Users need to decide which display server to run, and that means not being able to simultaneously use applications written for the other server - while toolkits may abstract this away most of the time, it's not guaranteed there won't be display server specific applications. I could probably come up with more.

You call the above a win? This "Mir is open source, so everything is fine" view is waaaaaaay too simplistic. Of course there's an easy way out, Intel already took it - simply don't accept Mir-specific code, let Canonical maintain patches downstream. And that's where the "private" part of dee's post comes in - if Canonical maintains all the stuff in their own island, it's all private to them, the wider community doesn't benefit in any way, only Ubuntu users do. Assuming of course that Canonical can handle the burden of maintaining support for Mir in all toolkits/drivers/whathaveyou, because if they don't do a good job of it Ubuntu users will get a buggy experience.

So no, we do not all win from the existence Mir. There's plenty of downsides and pretty much no upsides.

[V10lator]

- SNIP -

While all you say is true you forgot one important thing: Developers that don't want to support 2 display servers in their code but still want to let users decide which one to use will stick to X11 as it's supported through XWayland and XMir. So Mir also slows down the adaption of Wayland (and Mir itself).