Mir + Unity QML + Unity APIs = Unity

March 4, 2013 – 10:44 am

It all sounds good in theory…

Not too long ago, Mark communicated the vision for Ubuntu and Unity for 2013 as “[...] Unity in 2013 will be all about mobile – bringing Ubuntu to phones and tablets [...]” and my team is responsible for taking Unity to these hardware platforms.

What you should expect to see during this year is an overhaul of Unity in order to power a wide variety of display sizes (think phone to tablet to desktop to TV to…), input methods (touch screens & on screen keyboards, traditional keyboards & pointer devices, voice, and whatever else Tony Stark makes us think of [*]), CPUs & GPUs, external peripherals and everything else we expect from a modern OS.

Looking closer at the problem ambitious goal, we had to take a few interesting decisions how we possibly would get to where we want to be.

Today Unity (the rendering part of it) runs as a plugin in Compiz which sits on top of X and is a recurring source of frustration on the developer-, design- and finally also on the user side as a lot of our ideas (handling a multi-monitor setup in 12.04 or our plans for menu bars come to mind) require quite some intrusive changes to the underlying system. Regardless of how carefully crafted the solution, it is bending a stack to something that it wasn’t necessarily designed for. Bend over backwards too much and you will fall on your back, bugging users with nasty bugs, regressions, unexpected behavior and plenty more that drives the frustration level up.

[*] I’d really love to see motion gestures, accelerometer and camera based ones- volunteers welcome :)

Enter Mir

While evaluating our options, looking at extending the current stack to our needs, using the Wayland protocol (or any of its implementations) and comparing that with our designs & ideas we concluded that neither approach would allow us to do what we want in the quality that we would like to see for Ubuntu & Unity (at cost and in time).

Plus, there is the rather sizable challenge of pulling Wayland/X onto a mobile device, working with SoCs on driver support, tuning this stack for power consumption and performance and dealing with other issues of a stack that hasn’t been designed for a convergence setup as we envision it. A lot of distraction from the actual goal, to provide an outstanding experience across all the supported devices – from consumer electronics to desktop computing devices to enterprise devices.

The chosen approach was to develop Mir, our own Display Server which is engineered driven by the designs and requirements that our larger vision dictates – no compromises, no crude hacks, fully testable & tested, performance in mind, support for legacy X applications, developed by Ubuntu for Ubuntu. We have put together our thoughts in the Unity & Mir specs and will introduce the topic for discussion at UDS this week.

In accordance with Ubuntu Touch’s SDK we have also decided to transition the Unity Shell from Nux towards a QML based implementation. The new proposed monthly release cadence would allow us to iterate quickly and we are planning to tackle the transition component by component. Having one single code base that serves all variants of input & display types will help greatly to deliver the unified User Experience we are committing to.
Unity’s APIs are at the heart of your (Desktop) Shell and are constantly being evaluated and improved. In 2013 we will be looking at these APIs with a specific focus on resource efficiency, performance and actual API design to help increase the adoption with 3rd party developers.

Thomas Voss is giving you some more insight into Mir over at his blog.

…but in practice

Despite the large scope of our grand plan, it still looks good in practice and I am very excited about the change that we are going to drive.

While criticized as being full of mocks, placeholders and the like, all the code in Ubuntu Touch is what we deem production code and we have planned to remove stubs & mocks in the next little bit. A look at the published sources and following the development will show you our commitment to deliver high quality software. For Unity, this means there is one code base that has already demoed strong with the industry at CES and MWC and will be build upon – there is no distinction between demo & production code, in other words: what you see is the real thing and it’s improving every day :)

As Ubuntu Touch shows, Qt/QML has proven to be a rapid development “framework” that allows us to implement our user experience and more importantly to iterate over existing implementations with feedback from User Research and other drivers. The already mentioned proposed shorter release cycles will help us to land code faster and in better quality.

The “100 Scopes project” which was initially targeted for 13.04 is the first initiative to review our APIs and is well on it’s way – it is continuing to show us potential limitations of the current design  which are going to be discussed at upcoming UDSes.

Mir, despite being released only now has had some internal head start and is in certain areas progressing very well, to the extent that we are confident for Mir to start the discussion about landing Mir in an Ubuntu release soon.

Now, I am not trying to understate the problem we are will be solving and I can share with you that there are still some nights with less sleep (not quite sleepless though) about the unknown unknowns that are ahead of us.

I am certain that we will hit limitations in QML sooner rather than later which will have us look more closely into Qt and work with the large Qt community to address these.

Mir is an ambitious project with lots of parameters to it and we appreciate that we are trying to replace a stack that was maturing for almost 3 decades (no pressure, team ;).

Finally, designing a user experience that adopts itself seaminglessly between all the various devices and combinations thereof is no easy task and there are plenty of discussions how to close the gap between the Unity that you know today (up to Ubuntu Raring) and what we showed as part of Ubuntu Touch.

If it was easy it wouldn’t be fun! (famous last words)

2D or not 2D (or: learn, adapt, win)

12.04 shipped “Unity” (Nux) and “Unity 2D” (Qt/QML) , in 12.10 “Unity 2D” (Qt/QML) was deprecated in favor of “Unity” (Nux) and now Unity will be moved to Qt/QML? What the…

One thing I enjoy about Canonical is that we are not shy in providing leadership and will take controversial decisions if and when we believe they will help Ubuntu as a project to achieve the vision. Success however is only reached if you don’t rest on a made-decision but constantly evaluate where your decisions took you and what is happening around you. If needed you will have to go back and adapt to something new. A very basic concept of evolution in my books.

Back at UDS Q and prior to that, we discussed in length where Unity should be headed. Are we betting on Qt/QML, which at that time was going through a hair-raising transition or should we play safe and keep full control of what is dear to us? There were lots of technical evaluations, measurements, discussions, arguments and everything else you would do to make an influential decision like that. The outcome was that the implementation of Unity known as Unity 2D would be deprecated in order to not having to maintain 2 code bases with differing levels of feature completion, bug incompatibility and so on.

Fast forward 9 months, we have learned that our concerns have not been justified, Qt 5 has come out nicely and we have full confidence in using it as one core technology in our offerings.

Invitation to participate

Starting today we are opening up the development of this new exciting next step of Unity and invite interested developers to participate.

We will run sessions regarding Mir and Unity at this weeks UDS and would be happy to have you there. If you can’t make it to UDS this time then feel free to reach out to the team on the unity-dev mailing list (mailto: unity-dev@lists.launchpad.net). You can of course also find us on IRC (freenode, #ubuntu-unity).