Collaboration between ubports and yunit projects.

A Former User · 4 Oct 2017, 11:31

Hi! Sorry for my late response, but the last couple of days were a mess trying to setup yunit project. So here we go:

Based on preliminary discussions we had we both agree that moving from mir to wayland is the direction we need to follow. Having that base common goal in mind, the reasonable conclusion is that we may not need to have two different separate repositories and that this will make developers' lives easier. On the other hand I believe that two different projects working with the same repo will create a real mess regarding management, decision making etc.

In any case, I would personally prefer to have a common repo and figure out all the rest regarding management, decision making etc, rather than the oposite, as two different repos with (hopefully) the same code, will be very confusing to the outsiders. In any case, I would love to hear your thoughts/plans and suggestions as you are more experienced in the subject (unity, ubuntu, canonical, etc) than we are.

PS: Per my understanding before Canonical's announcement to discontinue unity and mir you didn't have any repo, but your people were contributing directly to Canonical's repo. Am I right? If this is indeed the case, then your experience working with Canonical will be a valuable asset.

flohack · 9 Apr 2017, 10:50

@jsalatas Well AFAIK we used custom repos for the device, kernel and vendor parts. All other stuff comes from CM, AOSP and some Canonical repos that are needed for building. But no one of us (again correct me if I am wrong) did contribute to unity or the system image.

Iorek · 9 Apr 2017, 18:07

The best way might be to have a mobile branch, a desktop branch, and a common branch ?
That way each group can try stuff out without worrying about breaking things for the other folks, and then integrate it to the main branch when it's stable enough.
It might even be worth it to have someone on each side try the code out before it gets pushed onto the main branch and then pulled into the other project's.
Of course that means we add an interface, and that might delay progress, but we're probably going to have to do that kind of things internally anyway, so why not do it also between the 2 projects ?

A Former User · 9 Apr 2017, 18:27

@Flohack said in Collaboration between ubports and yunit projects.:

But no one of us (again correct me if I am wrong) did contribute to unity

So maybe we could keep it that way? Of course you will have a saying in any development related decisions and also any contributions/suggestions/modifications you might need are more than welcomed!

flohack · 9 Apr 2017, 18:27

@jsalatas that need to be answered by the @Developers

WebDrake · 9 Apr 2017, 19:49

@jsalatas said in Collaboration between ubports and yunit projects.:

Based on preliminary discussions we had we both agree that moving from mir to wayland is the direction we need to follow.

Out of curiosity, why?

According to the thread here, there is going to be ongoing investment in mir:
https://plus.google.com/+MarkShuttleworthCanonical/posts/7LYubpaHUHH

... so why attempt to move away from this piece of the stack unless it's provably necessary? It could allow for some interesting points of collaboration with Canonical's ongoing IoT efforts.

A Former User · 9 Apr 2017, 19:51

@WebDrake It doesn't mention something about mir there $:\$

WebDrake · 9 Apr 2017, 19:55

@jsalatas said in Collaboration between ubports and yunit projects.:

@WebDrake It doesn't mention something about mir there $:\$

Yes it does. I said the thread, not the post

The particular comment in question (it's some way down in the discussion, I should have mentioned that) was highlighted in a later Reddit thread:
https://www.reddit.com/r/linux/comments/646mv8/mark_shuttleworth_i_came_to_be_disgusted_with_the/

Specifically, Mark Shuttleworth writes:

we have lots of IoT projects using Mir as a compositor so that code continues to receive investment. I agree, it's a very fast, clean and powerful graphics composition engine, and smart people love it for that.

A Former User · 9 Apr 2017, 20:03

@WebDrake Hmmm... interesting!

Thinking out loud: will that be a good idea to continue relying on canonical? Maybe they will decide to drop it and move to something else after a year or so. $:\$

WebDrake · 9 Apr 2017, 20:14

@jsalatas said in Collaboration between ubports and yunit projects.:

Thinking out loud: will that be a good idea to continue relying on canonical? Maybe they will decide to drop it and move to something else after a year or so. $:\$

Well, what I'd suggest to you is that this is an argument from theory and speculation. The simplest thing is probably to actually talk to the folks at Canonical and ask to what extent you can engage on this. If they say, "No, definitely go with Wayland", there's your answer. But if they are interested in having you continue to use mir, and are prepared to take patches from you in support of your work, then that potentially takes a whole body of effort out of your hands for the foreseeable future.

If in a year or two you find yourself having to transition after all, then you'll be facing that challenge from the point of view of having a lot more experience of the collective codebase and how it all works. But in the meantime, why create work for yourselves until you know you have to do that work?

A Former User · 9 Apr 2017, 20:21

@WebDrake Actually I would say that my speculation is based on actual facts. What would happen for example if we asked them about unity8 and convergence let's say 3 months ago? They would assure us that everything is set up for 18.04 and stuff like that.
It is obvious to me that whatever they say now about mir xan probably change in the near future.

In any case can we stick to the collaboration between yunit and UBPorts in this topic? And maybe have a different one to discus the mir vs wayland think?

WebDrake · 9 Apr 2017, 20:37

@jsalatas said in Collaboration between ubports and yunit projects.:

In any case can we stick to the collaboration between yunit and UBPorts in this topic? And maybe have a different one to discus the mir vs wayland think?

Yes, fair enough. FWIW I wasn't trying to launch a mir-vs.-wayland discussion per se. My concern was more that in creating the yunit/UBPorts collaboration, it's important to have a clear picture of what's genuinely needed for things to work.

I want this collaboration to be a big success, and with that in mind I think it matters to very rigorously ask the question, "What's the minimum amount of work we need to do to get this codebase ready for a working desktop?"

cornelcasanova · 10 Apr 2017, 04:41

@WebDrake I like the idea of keeping Mir alive

A Former User · 10 Apr 2017, 04:44

@cornelcasanova said in Collaboration between ubports and yunit projects.:

@WebDrake I like the idea of keeping Mir alive

we will not keep it alive. Canonical will do. Until further notice

Aury88 · 29 Apr 2017, 07:06

sorry for the intrusion
@WebDrake about the Mir support by canonical I think anyway they would continue to support it but limited to the IoT systems..so probably there will be not PC graphic drivers port/support. imho the transition to wayland is necessary even if canonical will continue to support it..that support probably will not cover the pc market

sverzegnassi · 6 Jun 2017, 14:34

@Aury88 The problem here is not whether or not doing such transition, but how.

Unity8 is a complex software (~110k lines of code) which has a huge number of dependencies, from the Ubuntu UI Toolkit to GLib, from LightDM to LibIndicator and AccountServices, and all the other components specifically written for U8: Scopes library, Ubuntu Download Manager, Ubuntu Keyboard, MediaScanner2, MediaHub, Thumbnailer, and so on...

Mir is just one piece of the puzzle. Before touching any single line of code, it's important to know every single dependency of Unity 8, and be sure that all the building infrastructure works properly (i.e. being able to successfully run all the test suites of these components).

What really concerns me is not moving from Mir to Wayland, but getting a proper estimation of the development costs.

I'm not sure that going forward with the Canonical idea of convergence ("One OS running on every device") is the best way to get things done.

I have the impression that having a single shell on different devices (phones, tablets and desktops) adds a lot of complexity to a single software, and Canonical probably got stuck on that.

I haven't seen yet any other project trying to do the same: Microsoft decided to develop two different OS's: only "Universal apps" and the UI Toolkit are shared between the two platforms; KDE is doing the same, developing Plasma Mobile and a convergent UI Toolkit called Kirigami; and ChromeOS is not Android, but it's able to run its applications.

Also, there's another fact I haven't seen any mention yet: moving Ubuntu Touch to Xenial would actually break any compatibility with the apps and scopes currently available on the Ubuntu/Open Store. Canonical had an hard time (since 2015) to figure out how to do such transition, and they never moved from Vivid (i.e. Ubuntu 15.04) as base for their Ubuntu Touch images.

refs.

Ads20000 · 6 Jun 2017, 14:34

@sverzegnassi I'm sure people would be willing to donate money to the project if you keep the convergence vision and otherwise they may not. I love this vision and when I have money I will contribute and I will definitely help test Unity 8 on desktop too!
Here's one person ready to pledge cash via Patreon if you commit to developing Unity 8 (on desktop, presumably): https://disqus.com/home/discussion/omgubuntu/make_gnome_shell_look_exactly_like_unity_7_with_this_theme/#comment-3344454071

sverzegnassi · 6 Jun 2017, 15:17

@Ads20000 A month ago, when I wrote it, many things were still unclear.
There were less people involved on the project, much less interest from the community, and the future of UT (and yunit) was much uncertain. During the last month, many people (even ex Canonical developers) jumped in, and many things got sort out. Also, the hype within the community increased a lot.

If you ask me what I think about convergence, I still think that it shouldn't be the first priority, since what has been left by Canonical still requires a lot of work. I still have some doubt that one single shell (which relies on many technologies written exclusively for Ubuntu Touch - e.g. the app confinement story, in particular) could satisfy the needs of both phones and desktops.

The question I ask myself a month ago was: could we expect people will still have interest on Unity 8 e.g. in an year? My answer now is that we can still put some effort on convergence, since now more people are involved on both projects, and keep everything alive.

Money and donations surely help, but I suspect there's still some trade-off. If you can't get developers to work on other parts of the platform too, it's hard to get things done. However this is not the case anymore

Aury88 · 2 Jul 2017, 06:15

Hi guys! about this topic I didn't read any news about the collaboration between ubport and yunit team and I'm curious to know if it is already reality or a work in progress.
For what I understand yunit has successfully ported unity code on debian stable and is backporting to 16.04. will this migration to 16.04 be useful for ubport too?
is it still possible to merge all the fixes and update both projects had done on their branch?

Also I would like to know if there is some sort of collaboration between ubport (and yunit) with the canonical mir team

about the mir support by canonical this is the actual situation
Also mate team seems interested in a mir-based wayland compositor/client. is the collaboration between all those project on this common objective possible and already started?
I hope in some news in the next Community Update

bye guys and thank you!