About the Mir display server

[mithion]

As some of you may have heard, Canonical, the company behind Ubuntu, has revealed that it has been developping a new display system called Mir to be used with their different flavors of their distribution instead of Wayland. Given that a transition to a new display system will probably happen soon for Mageia (maybe within the next version or two), the introduction of Mir may pose a dilema for the Linux community beyond Ubuntu. I was curious to see what any of you thought about Mir and Wayland and what this new display server means for Mageia. Of course, it's very early to make a call as nobody knows at this point how the driver ecosystem will evolve with both Mir and Wayland, but I have little doubt that with Ubuntu's large user base and general weight in the community, Mir will influence the ecosystem. What are your thoughts?

P.S.: I'm not a developper of any sorts and I'm asking purely out of curiousity as a Mageia user.

[digigold]

From what I understand, this is being designed with Unity in mind. I would think that when X is eventually replaced in Mageia it will be with Wayland.

Wayland points of interest:

#1 - Fedora has already indicated that they are moving toward Wayland. Remember MGA is a rpm based distro so I would think that they would be more likely to follow RH than Canonical/Ubuntu which is debian based.

#2 - KDE - Last January KWin’s main developer (Martin Gräßlin) started working for Blue Systems (Kubuntu & KDE dev co.) with one of the goals being a complete Wayland port. This means even Kubuntu will likely use Wayland.

#3 - GTK+ support is complete, except for client-side decorations.

#4 - Qt 5 has complete Wayland support, including the client-side decorations.

Most telling of all would be the fact that Mageia Media Sources already contain Wayland libraries, documentation and development tools!

[mithion]

I'm wondering though if this will fracture the linux ecosystem. Up until Mir showed up, every linux distribution had at least one thing in common, they all run X. Given it's already difficult to get good graphics support with one common platform, what will happen when developers presented with two platforms? Will one them become a second class citizen? Or is it feasible that both platforms will get good support?

[digigold]

Well, both Upstart and SystemD are supported pretty well. Although SystemD is far superior IMHO and came out a few years after upstart. But along with WIFI, graphics probably provide the most difficulties for the novice desktop user, so you may have a point. I also get the feeling that Mir may end up focusing on the mobile/lightweight side of things, but maybe I'm wrong. Although Canonical contributes to ChromeOS and ChromeBook's market share keeps increasing. On the other hand I believe Chrome Browser has been ported to Wayland so who knows.

TBH I tend to think Canonical's strength lies more in Marketing then development, but I've always been more of a RH guy myself so I'm fairly biased.

[doktor5000]

I'm wondering though if this will fracture the linux ecosystem.

Isn't this what makes up the linux ecosystem? To each his own? Diversity is king? ...
Thats basically the reason for hundreds of distributions, often only differing in small details. And forks of forks of forks. And so on ...

[mithion]

Isn't this what makes up the linux ecosystem? To each his own? Diversity is king? ...
Thats basically the reason for hundreds of distributions, often only differing in small details. And forks of forks of forks. And so on ...

I agree that choice and diversity are definitely one of the great things you can experience when you step into the world of Linux. However, a lot of the diversity comes at a higher level of functionality (e.g. rpm vs deb, or command line approach vs gui etc...). Most of the time, it isn't too disruptive to the basic functionality between the kernel and hardware. However, the display server is a pretty fundamental part of the user interaction, and by its very nature is complex to develop. It's been difficult over the years to have good driver support on Linux. I'm not familiar enough to know if the difficulty comes from the lack of man power which means good performance is possible but not enough people are developping it. Or if it comes from a bad design, in which case it will never work perfectly no matter how much money and people we throw at it.

I really don't mind diversity. In fact, if I had the ressources that canonical has, I'd probably do the same and just roll my own so I wouldn't have to compromise on the features that I want. However, Ubuntu doesn't exist in a vacuum and the introduction of Mir may definitely impact everyone, not just Ubuntu users. My ultimate hope is that both Wayland and Mir are extremely well designed, and are a breeze to code for. In that scenario, supporting both platforms wouldn't be a problem and everyone would be happy.

Diversity for the sake of diversity doesn't always end well. For example, look at the BSDs. I have a soft spot for them just because I like they idea of built-in-house toolchain+kernel which makes for a very consistent system. However there is no denying they are feeling some pain from all the change that are occuring on the GNU side which is making it more and more difficult for them to make use of the ecosystem of software available to us. This is an example of where the rolling your own approach has caused problems. Maybe I'm wrong in my analogy, but I just hope that Ubuntu doesn't become to the rest of Linux, what Linux is to the BSDs.

[doktor5000]

Well, ubuntu isn't everything. Most of the major changes and projects they introduced, lack behind their counterparts,
and didn't receive wide adoption. Take upstart, bazaar or Unity as examples.

For Mir itself and some clarification what is what (remember, Wayland is a definition of a protocol, and Mir is intended to be a display server) you might also want to take a look at https://wiki.ubuntu.com/MirSpec It seems also more tilored towards mobile devices (meaning smartphones/tablets).

You should also definitely take a look at https://plus.google.com/100409717163242 ... Dq6BAgdpkG (especially the comments)

I also didn't mean that diversity is necessarily a good thing. It's linux strenth and weakness or even achilles heel together at the same time.

[wobo]

Well, the difference between current systems is mostly userland diversity. Currently you can use KDE and Gnome on the same system, even use deb packages in a rpm environment (to some extent). But Mir and Wayland are beyond the current layer of diversity. Now we have one graphical system we will then have two of them, which means the group of developpers who are working on drivers and tools and whatever is related to display topics will be divided by 2, leaving only half of the manpower for each graphic system (in theory).

This is only the start. As with every other "fork" the different systems may drift apart until it has consequences on application development as well. Then the user community (private and commercial) will also be split in two - resulting in splitting the market share of Gnu/Linux in two groups.

I am a hardliner when it comes to principles of the Linux world (like diversity). But this may go beyond the current very diverse One World of Gnu/Linux.

All this is certainly thinking very far ahead and based on thoughts of a layman.

[mithion]

You should also definitely take a look at https://plus.google.com/100409717163242 ... Dq6BAgdpkG (especially the comments)

Is it at all possible that they may reverse ship. From this link you posted and other tidbits I've gathered around, doesn't sound like Mir is getting a lot of love. The KDE guys are moving along with Wayland and so is Team Gnome. If such major players reject Mir, Canonical will stand in isolation. Are they trying to become the next Apple, except using a Linux kernel instead of Darwin? I have huge doubts (and it seems most people have the same doubts I do) that they have the man power or expertise to build everything in-house.

This is only the start. As with every other "fork" the different systems may drift apart until it has consequences on application development as well. Then the user community (private and commercial) will also be split in two - resulting in splitting the market share of Gnu/Linux in two groups.

If Mir really takes off, these are my thoughts exactly. Both sides will suffer in quality of driver support. I still don't quite understand why Canonical is doing this except for the sole reason of having complete control over the intellectual property that is Mir. Because the reasoning they give in the specs for skipping Wayland don't make any sense. And even if they have legitimate issues with Wayland, it would be way easier to contact upstream and see what can be done since Wayland is still in under heavy development anyway.

[digigold]

...Canonical will stand in isolation...I have huge doubts (and it seems most people have the same doubts I do) that they have the man power or expertise to build everything in-house.

Also with regards to Mir development, they won't be able to "borrow" code from RH to use either..

[marast78]

What I see is Canonical is still not making any money. The 'Amazon ads' issue introduced in 12.10 didn't help much either. Though I didn't find it as such a big problem like many others. They didn't even inform the main developer of LightDM that they are going with Mir instead of Wayland, even though they knew it for about half a year, he got a bit angry about that. No wonder.
Then Mark says something like Unity was there sooner then Gnom3 and that they pioneered the 6 month release cycle. BTW, Adam responded to that on his blog: http://www.happyassassin.net/2013/03/11 ... the-truth/

But I really like Unity. Many of Gnom3's issues are not present here and it's generally quite usable. Unfortunately, after recent movement, I'm back on Fedora. Canonical (Mark?) is trying to push forward it's baby in not a very polite manner, so to speak. I don't like that. Canonical needs profit, but with recent affairs, it will hardly come.

[mithion]

Yeah I really don't get the introduction of Mir. There's already a great example of a succesful commercial venture using open source, it's called Red Hat. Red Hat does a lot of development with enterprise capability in mind for themselves, but a lot of it contributed to already existing projects and made available to everyone else. And whatever the community introduces, Red Hat benefits also. By rolling your own, you're basically cutting yourself off from everybody. The only thing that worries is that Ubuntu is a heavyweight. They have the most users (of the order of 20 million, possibly more I believe) and so it's not unreasonable to think they may sway major players into shifting developers for Mir just because of their large user base. It's hard to tell who has the big end of the stick right now. Many have spoken against Mir already, but Mir isn't schedule to be finalized until Ubuntu 14.04 so a lot can happen in a year.

[marast78]

Ubuntu is losing users in a rather constant rate. And if they don't wake up, it will get bad.
Right now, Intel is developing drivers for Wayland and Qt5 and GTK3 ports to Wayland are practically complete as well.
Nvidia will likely support it too, as it needs Linux for those render farms used by industry. And Linux in case of Nvidia means RedHat. Not only in case of Nvidia though. And after Wayland is defacto X replacement, Intel's and Nvidia's drivers will ikely be 'just done' and fully usable. I don't know what Ati will do, but since FGLRX is still bugged as hell, it may not matter so much.
Intel support is being led by the Wayland creator and so Mir support is very unlikely.

Canonical thinking Mir will turn Intel or Nvidia to their side is just absurd. But it's their mistake, not ours to solve.