Canonical Posts 15 Mesa Patches To Support Mir

[Delgarde]

It would lead to canonical's mesa packages diverging more and more from upstream, potentially introducing issues.

So? That's not a problem for anyone but Canonical, and for the most part, it's a self-inflicted one. They're the ones choosing to go in a different direction from the one chosen by pretty much everyone else - both upstreams, and the other distros.

[mrugiero]

Ok, to be fair, it's still weekend, the patch set has not been posted too long ago.
There's still time to reply. TBH, I don't think this will happen.

Yes, I didn't considered that because I'm usually pretty lost with the days.

I fully agree with you. And to me a big chance for Canonical to get back into
discussion with the freedesktop.org devs is to show up and discuss and present something at XDC2013.
After all they [Canonical] want some of their code merged upstream to an fdo project [Mesa].

Of course, this needs balls... Don't think they'll do this.

Fully agree.

We DONT CARE if you like or hate Ubuntu, Fedora, Suse, Torvalds, Mir or Wayland. Shut the fuck up and stay on topic.

We don't care if you like reading about it.

Wayland is just RPC library, so I can even fly to space with wayland - if someone will imlement it. Weston is display server (just look at source code before saying that it isn't), it enforces CSD and if someone will create yet one wayland-based server, you will see even more critisism than with Mir: "oh, they give up CSD to make our Weston less popular because no one toolkit writer actually wants to implement CSD".

Yet, it's a reference, and nobody is supposed to actually use it. If KDE wants their compositor to do server side decorations (which, IIRC, they do want), they will do that. And the same for everyone.
Weston is not supposed to be popular. You ask me to read the code, I ask you to read the readme. It's not supposed to be used at all in a real desktop.
KDE implements its own compositor, E18 implements its own compositor, GNOME Shell AFAIK implements its own compositor. None of them use Weston.
The toolkits aren't built again Weston, they are built against Wayland, so they can't care less how Weston decorates its windows or avoids decorating them.

Weston is compositor so what the F are you talking about? Yet another ubuntard.

A compositor can be a server, according to the docs. Can be otherwise, too.

[duby229]

Thank You.

Whether devs want to admit it or not, there is a certain amount of end user choice involved. The choice to accept the dumbass shit you enforce, or the choice to move on to something else. At minimum there will always be that choice.

[mrugiero]

So? That's not a problem for anyone but Canonical, and for the most part, it's a self-inflicted one. They're the ones choosing to go in a different direction from the one chosen by pretty much everyone else - both upstreams, and the other distros.

It seems like I have to repeat this: no graphics programmer targets mesa. Any difference between OpenGL specification and what mesa does is a bug or a missing feature. So, if there is any difference, is a bug in the one who diverges from the specification, and this means that either:

A) Upstream's code has a bug, in which case they will probably share it with Canonical, and mesa will have to fix it, and Canonical will probably just backport the fix;
or
B) Canonical's code has a bug, in which case Canonical should fix it.

If it goes upstream, the same applies, with the exception that Canonical's bugs will bother also non-Ubuntu users.

EDIT: Not that I oppose to them being accepted, as I said, I want Mir to either fail horribly or to be a good product; what I don't want in the least is something in between, causing fragmentation AND being a bad product.

[intellivision]

It seems like I have to repeat this: no graphics programmer targets mesa. Any difference between OpenGL specification and what mesa does is a bug or a missing feature. So, if there is any difference, is a bug in the one who diverges from the specification, and this means that either:

A) Upstream's code has a bug, in which case they will probably share it with Canonical, and mesa will have to fix it, and Canonical will probably just backport the fix;
or
B) Canonical's code has a bug, in which case Canonical should fix it.

If it goes upstream, the same applies, with the exception that Canonical's bugs will bother also non-Ubuntu users.

EDIT: Not that I oppose to them being accepted, as I said, I want Mir to either fail horribly or to be a good product; what I don't want in the least is something in between, causing fragmentation AND being a bad product.

So you would actively denounce any distribution that would consider transitioning from Xorg to Mir instead of Wayland? Failing that, wanting to see a project fail is petty, very petty.

[mrugiero]

So you would actively denounce any distribution that would consider transitioning from Xorg to Mir instead of Wayland? Failing that, wanting to see a project fail is petty, very petty.

Didn't understand what you meant. Put simple, what I want is for it to be worth the fragmentation (or at least not completely unworthy of it), or not provoke it.
If the project is somewhere in between (not including) total failure and anything else, the harm (the fragmentation) will persist. So, if that's the case, I want it to be a good product, so at least the fragmentation is not completely in vain. But if the final product is half assed, I prefer that it doesn't cause fragmentation, which is only achieved if it doesn't even serve Ubuntu well. As I already said, it's an 'or' condition.

EDIT: Also, if any other distro decides Mir fits them, I want the patches to be merged. If not, it doesn't make any difference for Canonical, and the only it can make for anyone else is detrimental (because the only feature they will see while not using Mir are possible bugs in shared code), so in that case I'd prefer the patches to be rejected. Since it's too early to expect any other distro to go on wagon, I think the sanest compromise would be to make an upstream branch for Mir support to ease upgrading mesa for Canonical without messing with other users.

EDIT 2: I seriously can't see why I would denounce anyone using Mir. What I want has nothing to do with denouncing anything. I'm fully aware I don't have (nor want) any authority on what distros does. I just stated how I'd want it to end up, which isn't necessarily failing. If it doesn't fail, I'd actually prefer mass adoption, since as long as only one is chosen, fragmentation is limited. But I tend to think with all the work other projects already did with Wayland there's no chance of Wayland being abandoned by everyone else, and that means any degree of success of Mir will cause fragmentation. Then, again, I'd want it to have epic success, to worth the fragmentation.

[doom_Oo7]

At this point we are pretty much dependent on Canonical (or maybe Google) giving us a decent Linux desktop because not a single one of the rest of them have a clue. Except maybe Mint... but those guys are heavily basing on Ubuntu so that explains their sanity.

??? What would be your problem with debian for desktop use for instance ? It works quite well, and is not lacking 6 years behind like it was ten years ago.

[dee.]

So you would actively denounce any distribution that would consider transitioning from Xorg to Mir instead of Wayland? Failing that, wanting to see a project fail is petty, very petty.

Well, the happy news is, no distro is ever going to "consider transitioning to Mir" for the simple reason that no desktop besides Unity supports Mir - Unity and Mir are fully codependent, so to speak. Now you could say, maybe some other distro adopts Unity and runs it on Mir. Well, Canonical has made it extremely hard for anyone else to use Unity - even the Arch guys have trouble getting it running. Not only that, but Unity is partially dependent on proprietary code - the "smartscope" serverside software is proprietary, meaning that no one else will be able to replicate that functionality, without relying on Canonical's proprietary servers.

So in order for any distribution to be able to "transition to Mir", there would first have to be a DE willing to transition to Mir. What DE's are there? KDE, GNOME, Xfce, LXDE/Razor-Qt, E, MATE, Cinnamon. Of these, KDE, GNOME, Xfce, LXDE/Razor-Qt and E have publicly declared their support for Wayland, and that they have no plans on even considering implementing Mir support at this time. Cinnamon and MATE are both developed (or heavily influenced) by the Mint team, which while hasn't explicitly stated their plans regarding Wayland, has at least implied as much - IIRC, Clem has stated in a blog post that they will not change their plans because of Mir.

It's very unlikely that any other DE will ever run natively on Mir, for a couple of reasons: no one really has any interest in running Mir apart from Canonical, Canonical has not at this time presented a stable protocol or API that is guaranteed to stay stable, and it is unclear if they ever will do so, and Canonical is developing Mir for the needs of Unity only - no other DE would have any option to influence the development, ie. all other DE's would at least be 2nd class citizens on Mir. On the other hand, none of this applies to Wayland, which is open for any DE to join in and contribute. Any sane desktop is going to weigh the benefits and see that Wayland is the better option for them.

As for running a whole DE on XMir, that's just incredibly stupid - as would be running a whole DE on XWayland. It's unavoidably slower than plain X, adds a new layer of vulnerability, with none of the benefits of the new display server. What kind of idiot would even want to do that...

[nomadewolf]

Wayland is just RPC library, so I can even fly to space with wayland - if someone will imlement it. Weston is display server (just look at source code before saying that it isn't), it enforces CSD and if someone will create yet one wayland-based server, you will see even more critisism than with Mir: "oh, they give up CSD to make our Weston less popular because no one toolkit writer actually wants to implement CSD".

I think what needs to be stressed is that Weston is a reference compositor. It's how the wayland team thinks compositors working on the wayland display server sould be implemented.
Anyone can use Weston, forkit or implement something else entirely.
If they get criticised for this or that choice, will cross that bridge when we get there.

[nomadewolf]

That guy sounds like a real pompous ass. So basically your typical Linux developer.

At this point we are pretty much dependent on Canonical (or maybe Google) giving us a decent Linux desktop because not a single one of the rest of them have a clue. Except maybe Mint... but those guys are heavily basing on Ubuntu so that explains their sanity.

At least with Mint, their choices are ALWAYS made with the users in mind.
So if the user base doesn't want MIR, it won't be shoved down their throats.
Also, Mint has a 100% Debian based version, just in case...