Posted on October 5, 2017

A Plasma Phone
In one of my latest blogs, I’ve explained what convergence is, how Plasma benefits from it, and why we consider it a goal for Plasma. This time around, I’ll explain the how, how it works across the stack and how we implemented it. Naturally, this article dives a lot deeper, technically, than my previous one.

Convergence plays a role at different levels of the whole software stack. In this, more technical article, I’ll look at different layers of the software stack, from boot/kernel and middleware to UI controls and overall layout and input methods. After reading this article, you’ll understand how Plasma allows to use the same software on a range of devices, which parts are different, and where code sharing makes sense, and thus happens.
Keep in mind that Convergence, at least for Plasma, doesn’t mean that we ship a lowest-common denominator UI so it “kind of” runs on all things computer, but that it provides a toolbox to build customized UIs that allow taking advantage of specific characteristics of a given target device.

Lower Levels and Packaging

Plasma — same code, different devices
One aspect of convergence that is of course the deployment side. This doesn’t just include the kernel and bootloader, which needs to be compiled differently for ARM devices and for x86 devices. The rest of the stack is by now largely the same. We are now using the same set of packages and CI for both, mobile and desktop builds, in fact most packages are the same, and the difference between a device set up for mobile use cases and desktop is the selection of packages, and what gets started by default. Everything is integrated to a very large degree, lots of work is shared which means timely updates across the device spectrum we serve.

Controls

Plasma Mobile
When it comes to user interface controls, such as buttons, text fields, etc., convergence is mostly a solved problem. Touch input is possible, Qt nowadays even ships a virtual keyboard (which Plasma uses for example for password input in the lock-screen), and buttons react to touch events as well. QtQuick-based user interfaces often work quite well with both, keyboard/mouse and touch input, in fact touch is one of the design goals of QtQuick.

Not everything is perfect yet, however, especially text selection and keyboard control of QtQuick-based UIs often still requires custom-code, meaning it needs more development and maintainance time to get right. QWidget-based UIs are still a bit ahead of the game here, though often the benefits of also being able to deploy an app on touch devices (such as many Android devices out there!) make QtQuick an attractive technology to use. We see more and more QtQuick-based applications, as this technology matures also for desktop use-cases.

Widgets

Convergent Plasma Widgets
Plasma is made of widgets. Even in a standard Plasma desktop, everything is a widget: The menu in the bottom left is a widget, the task manager is, the system tray is a widget, and there are widgets inside the system tray for notifications, battery, sound, network, etc.. Plasma is widgets.
These widgets can be used on any device of course, but it doesn’t always make sense. Some of these widgets are very specific for desktops. The task-manager (that thing you use to switch windows, which is usually located in center of the bottom panel) doesn’t really make sense on a mobile device. For a mobile device, which needs larger areas to touch, something more aking to a full-screen window switcher is useful (and in fact what we use for Plasma Mobile). Other widgets, such as the network connections widget or battery and brightness widgets are perfectly suitable also for mobile devices. Plasma’s architecture allows us to re-use the components that need no or just little changes and use them across devices. That means we can concentrate on the missing bits for each device, and that in turn means we can deliver a feature-rich and consistent UI across devices much easier, while making sure the specific characteristics of a given form-factor are used to their fullest extent.
Again, by sharing the components that make sense to share, we can deliver higher quality features for a given devices with less effort, and thus quicker.

Shell and Look & Feel

Look & Feel setup for Plasma
Plasma can dynamically load different so-called shell packages. The shell package defines the overall layout of the workspace environment. On the desktop, it says that there’s a fullscreen wallpaper background, with a folderview, a panel at the bottom and the widgets which are loaded into that panel: application launcher, task-manager, system tray and clock for example.
The shell package is different for each device, as this defines the overall workflow, which is highly dependent on the type of device.

To take differences between devices even further, Plasma has the concept of “Look and Feel” packages, which allow further specilialization how a device, well, looks and feels. There’s the widget style and the wallpaper of course. The Look and feel package also defines interaction patterns, such as if a settings interface should use “instant apply” when a setting is changed, or if it should present an “Apply and Okay” button for the user to save settings. Mobile devices typically use instant apply, while desktop interfaces (at least Plasma’s) use the “Apply and Okay” concept throughout. For Plasma UIs, this can be changed dynamically. Plasma’s Look and Feel features is not just useful in the convergence aspect, it allows also for example to switch between a traditional default Plasma setup and a workspace that closely resembles Unity. These “Look and Feel” packages are available through the KDE store, so they’re easy to install and share. There’s even a cool tool that allows you to create your own Look and Feel packages, very much like themes.

Subsurface Mobile, Linux, Android and iOS from the same code-base
Finally, at application level, we see more and more convergent applications. Kirigami, a high-level toolkit that supplied components for consistent, touch- and keyboard/mouse-friendly application nagivation and layout makes it very easy to create applications with responsive UIs that adapt well to screen size and density and that show flexibility in their input methods. This doesn’t just work between Laptops and phones, but also allows to create one app that works equally well on desktops, laptops, phones and tablets.
Kirigami complements Plasma’s convergence feature on application side, and we recommend it for most newly developed apps. With Qt and QtQuick being a viable target for Android devices, it increases the possible target audience by a very large degree. As an example, Subsurface Mobile, an application for scuba divers, uses Kirigami and works on Linux desktops, Android and iOS, all from the same code-base.

Make it happen…

If you like the idea of convergence, why not join KDE and help us work on Plasma? Perhaps you’d love to see Plasma on a mobile phone? In that case, consider backing the crowdfunding campaign for the librem5 so we can build a convergent phone!

3 thoughts on “Plasma Convergence, technically”

1. Looks great. Seems to be doing what UWP set out to do and failed as it didn’t catch on. Hoping this fairs better.

2. Kurt October 8, 201721:20

   This is great! Hopefully the UBports team will support the Purism Librem 5 as well, the more (quality!) choices we have the better for everyone!

3. John October 8, 201723:34

   What would you think of some generic approach to (automatically?) transform desktop UIs to mobile UIs?

   The Menu and Toolbar items (if the latter ones only contain redundant entries, they could be left out entirely) would go to the left drawer. Another possibility would be to define a special toolbar that contains the elements the left drawer should show.

   The main area would be used for the content.

   The right drawer can contain the dockers (or whatever areas there are at the edges of the desktop app), maybe as collapsible items. Alternatively it would be an idea to have the docks as drawers on the edges where they would be at the desktop app, then the menu would have to go somewhere else, maybe the title bar.

   Tabs could just remain tabs, same goes for context menus (appear on tap and hold).

   The resulting UI would maybe not be perfectly suitable for every application, but I could imagine this for some core apps like Dolphin, Kate, Okular or Gwenview.
   This could result in mobile apps (at least for tablets) with the complete set of features the desktop apps have, without the need to write any new code, not even for the UI.

Comments are closed.