can you do some power usage tests and AMD or Nvidia?While Linux enthusiasts are quick to criticize Canonical's Mir display server for a number of reasons, the performance impact of having XMir involved in the graphics stack is no longer much of a legitimate complaint when using the very latest software -- at least for Intel users; the 2D and 3D performance is both generally in good shape.
There are valid reasons for hating Unity, Mir, X.org, wayland, Canonical, microsoft, Nokia, Google, Apple, Valve, Oracle, Red Hat, etc.
Everybody loves to back the small guy.
There are also awesome reasons to love these evil behemoths.
I love Canonical. I do believe they are evil in their own way. I feel it is a neccesary evil. Canonical almost singlehandedly created the Linux revolution. Before Ubuntu 5.4 there were a few geeks and professionals who new about linux, now everybody can use linux. (I actually believe strongly that Linux is easier to use than Microsoft when compared correctly)
Xmir is a stepping stone, simply a temporary layer needed for compatibility to ensure that there are working software to use on Mir. Because existing applications / toolsets needs to be re-built to bind to the new back-end. Eventually most if not all software will be built on the native Mir / Wayland enabled toolsets. In the mean time we need the translation layers. Refering to Xmir in any arguments before/against Mir is irrelevant.
Mir represents a promise: To make your mobile handset / tablet / etc run "stock" Linux applications. I have not heard the same objective/purpose/dream mentioned for Wayland, but I know that Wayland works on Rapberry Pi. That's good enough for me. In the mean time huge amounts of commercial development effort goes into this. Google isn't going to open up their tool set any time soon. So we need Mir.
I'll be running Wayland on my system. Users realy don't care whether they have a protocol delivered by an implementation or an implementation delivered by a protocol. Users just want to install their programs and be happy. I am happy with my Kubuntu and their commitment to Wayland. If Kubuntu disappears I will need to invest (heavily - it gets harder and harder as you get older) into switching, probably to OpenSuse. Must love their installer. But I digress.
The problem that I have an issue with is Fragmentation. Is there any other major distributions that will include Mir? Will Mir developers need to fight with Gnome-wayland to get their pre-requisite fixes ported to upstream? Will Gnome itself be forker? Or will it become Unity-on-Mir-sans-Gnome? Will Mir-GTK remain compatible with Gnome-GTK? Will GTK be forked into a new ever increasingly less compatible API?
Fragmentation spreads already rare development resources. My question is: Why do we need two platforms: Mir and Wayland?
Competition. Without something different pushing progress forward, you get stagnation in development. Let's face it: Development happens fastest when ripping off other peoples good ideas. But if there's only one person making ideas, you never have anything else to push development in other directions.Fragmentation spreads already rare development resources. My question is: Why do we need two platforms: Mir and Wayland?
Or putting it another way: Why do we need Gnome, Unity, KDE, and Xfce? Lets just all go with Gnome 3 to maximize use of resources. Of wait, you hate Gnome? Too bad, its your only option.
But you see, open source means you can contribute your own ideas to other people's ideas, because they are not strictly behind the closed doors of commercial secrecy, so there is never "only one person making ideas". And if you think it wouldn't get accepted, at least try first, because fragmentation brings problems with it.
I think there are fundamental differences between Mir and Wayland (like compositor) and the "extension" way Wayland responded to Mir's features seams like another Xorg situation.
Plus that is clear now how much behind would Ubuntu touch been if they where waiting for Wayland to work the way they wanted.
Let them (both projects) show what they are capable of and then history will judge them fair.
Last edited by verde; 09-25-2013 at 09:06 PM.
When it comes to toolkits, display servers, API providing infrastructure in general, this means two or more different ecosystems need to be maintained. In the display server front it is partially (almost totally, thankfully, but there are still exceptions) alleviated by the existence of toolkits, which can abstract the infrastructure to most apps.
If you think there are fundamental differences, I would like to know about them. The most important I know about is server side allocation versus client side on Weston, but that's not mandated by the protocol, is just how the reference compositor works.I think there are fundamental differences between Mir and Wayland (like compositor) and the "extension" way Wayland responded to Mir's features seams like another Xorg situation.
How much?Plus that is clear now how much behind would Ubuntu touch been if they where waiting for Wayland to work the way they wanted.
Wouldn't have helped, maybe, to express their concerns before developing their own? I'm not saying it would be all rose, but they didn't even try to collaborate with upstream, and that's a bit annoying to me. Nobody should expect others to read their mind.
I really don't see the problem with multiple display servers. In the end I am not willing to believe that Canonical decided to spend resources, receive never ending bashing and risk their existence just because they don't want to collaborate or because they are evil.
It is the same old story like Unity. And in the end Unity became a masterpiece of speed and productivity...
I don't say I like or not Mir. I want to see it. I didn't see any fragmentation or lower development pace for Wayland yet, so.. Don't forget XBMC was ported to Wayland and Mir in a week.
Last edited by verde; 09-25-2013 at 09:44 PM.
I'm not stating such a thing. They didn't state any valid technical reasons that I know of and that weren't refuted. They might be evil, as much as any other company that just cares about their money. Red Hat is such another, and Intel the same. I happen to agree with some of the things they are doing right now, but that doesn't blind me to know they do it because it's good business for them and nothing else.I really don't see the problem with multiple display servers. In the end I am not willing to believe that Canonical decided to spend resources, receive never ending bashing and risk their existence just because they don't want to collaborate or because they are evil.
However, there is the matter of control, which is the only speculation I can come up with, based on the few things that I got to know (for example, that most technical reasons, at least the ones given on the announcement, were wrong). It could have also been a mistake. But if so, they are just making it worse by not admitting it and stopping the resource waste. Business-wise, the control they get might be good business, but it might also be a mistake, since carriers could just take Wayland for free if they want.
It is not the same old story as Unity, for a simple reason: you might dislike the flavor, but the only fragmentation it introduced are the resources Canonical invest on it. It doesn't bring problems to anyone else, nor extra work. The only valid reason to complain is that it doesn't suit you, and that's easily solved by not using it. I don't extremely dislike it (I use it at college), but it's far from being my favorite. I like better the traditional desktop metaphor. I'm aware such a thing is a matter of personal taste, though. Mir and Wayland are a different story, it's closer to toolkits conflicts than to DE conflicts, and you should be aware of that.It is the same old story like Unity. And in the end Unity became a masterpiece of speed and productivity...
I'm not bashing Mir to the ground, either. The only thing I stubbornly dislike is XMir for running desktops, and even though I do (I have my reasons, which I won't discuss if you don't want me to), I admit most of the problems it could have brought are solved by now; the only remaining I can think of, aside for the current lack of blobs support (not a problem that could remotely affect me), is the possible confusion on who to blame for bugs; this one will probably mean upstream will not check possible bugs on X if they appear in Ubuntu, until further confirmation by someone running stand alone X. I'm just tired of the denial of the problems that could arise because of such a choice to create Mir, and I think it was reckless to create it before trying to go the upstream way. The way I see it, you need a very good reason to take the risk of fragmenting the desktop, and I haven't seen any technical one. Remark the "risk" word, because I don't think it's something that will inevitably happen, but that is possible.I don't say I like or not Mir. I want to see it. I didn't see any fragmentation or lower development pace for Wayland yet, so.. Don't forget XBMC was ported to Wayland and Mir in a week.
On the tastes front, my favorite DE is XFCE, so I will probably use the infrastructure I think is most advanced that it can run on, which right now and probably for the next year is X.org. Later, I think it will be Wayland, but there are no proofs to such idea.
EDIT: Also, the fragmentation will obviously not appear ON Wayland, but it would (if it does) appear on the apps relying directly on some part of the display infrastructure. As I said in another post, this is almost completely alleviated by toolkits, except for DEs. As long as nobody dares to talk directly to libwayland or libmirclient, the fragmentation will be reduced to the same introduced by DEs, which is just multiplied efforts on such projects. On the XBMC thing, I'm aware, but I didn't look through the code to check how it was done. If done through an abstraction layer, this means extra overhead. It's kind of like having the compatibility layer like I said before. Of course, such a thing would be needed to support at the same time X.org and Wayland, so the only difference would be when/if X.org support was dropped. If it were just Wayland, you would be able to drop the extra abstraction, reducing function calling overhead, while being there Mir and Wayland it will still be two different platforms on Linux to support, meaning the abstraction should stay there. Anyway, I think it's probably just minor overhead, I was just pointing it out.
Last edited by mrugiero; 09-25-2013 at 10:19 PM.