Creating an open device is a difficult challenge; the software is (mostly)
there, but the hardware is a different story. Aaron Seigo has been working
on the Vivaldi tablet as part of the Make Play Live effort and reported on
some of the hurdles that have been encountered trying to produce the device
at Akademy. There are lots of
pieces that go into such a device, so finding a combination that
works and can be sold is a non-trivial task.
touch-enabled version of KDE's Plasma environment—came out of a discussion that various people working on
Plasma had about the
technology and asking "where do we go from here?", Seigo said. KDE has a
desktop suite, with office, email, and many other applications, but "is that
all we want to do?". To he and others, it felt like KDE was treading
water, but the discussions made it clear that some in the project were not
happy with just that.
He believes very strongly in freedom and technology, and he started looking
beyond the desktop and laptops where KDE has traditionally been focused.
When you look at mobile devices, set top boxes, and other systems like those,
you don't see the freedom and openness that we have come to expect. There
is an inherent need for some humans to hack, but devices are
"increasingly not places where you can hack, unless Apple says you can".
There is this idea that a "tablet is a tablet, a laptop is a laptop", he
said, but that is "increasingly silly". There is a continuum of devices,
without sharp divisions between them. We have started to see others
picking up on that, and releasing hybrid devices recently, like a media
center that is controlled by a tablet or phone, tablets with keyboards, and
So, Plasma Active arose out of Plasma Netbook and Plasma Desktop. It
provides one KDE and Qt-based technology that can be used across all kinds
of devices. The difference between Plasma Active and the netbook/desktop
versions is 10-15,000 lines of code—out of a code base of some
third-of-a-million lines. So there are "tiny differences" between the two,
and things written for one will work on the other. This is a "compelling
reason" to use Plasma Active on all these different kinds of devices, Seigo said.
Android: best friend and worst enemy
Seigo cited Android as the "best friend and worst enemy of open devices".
It uses the Linux kernel, and it is great that there are so many devices
out there running Linux. But Google does no GPL enforcement, which results
in mostly binary-only devices. For device manufacturers, getting Android
to boot is the end goal so that the device can be sold. Once they can
deliver a working binary kernel to their customers, they are done.
We are dealing with cultural and business barriers when trying to deliver
open devices, he said. All the manufacturing for these devices is in
China, increasingly the design is being done there as well. To build
open devices, you must work with the cultures of Asia, but most KDE
developers are based in Europe or North America and are not familiar with
In addition, the manufacturers are "all about volume" and, at least so far,
open devices are not selling in quantities that make them interesting. He
is not just reporting these problems "because it sucks", he said, but
because "I think there are things we can do about it together".
He asked: Can we overcome these problems? His experience shows that it is
"a very big mountain to climb", but it is something that the community has
to take on itself. These problems are not something that big companies are
in, so "we need to take our destiny in our own hands".
From an implementation standpoint, there are three pieces of software that
need to work well together. The kernel, which needs a user space that
works with it, and a "human experience" on top of that. Seigo uses the term "human experience" rather than "user experience"
because "humans are not users", he said.
As a community, KDE
does the human experience part, and there are folks in the project with
some experience in the other two. Seigo asked how many in the room had
written a kernel module and got a few raised hands. "We need you guys", he
said, and asked that they bring their friends. These days, user space is
tightly coupled to the kernel, he said, so the two need to be in sync.
Plasma Active itself is ready to go to provide the human experience;
version 3 will be released in the next few months.
One recent addition is a "nice touch-friendly file manager", Seigo said,
and Plasma Active is more than just a desktop shell. It is enabling other
applications, like Calligra
and Marble, to work well on touch
addition, a recent two or three day effort turned Okular into a touch enabled
e-book reader using QML.
Lots of code has been taken from the desktop for Plasma Active, but there
are parts that will flow back to Plasma in the future as well. Many KDE
applications can be made touch-friendly relatively easily, he said, and
developers of any applications that might ever run on a tablet, phone, set top box,
etc. should be thinking about that. When he hears application developers
talking about separating the business logic from the presentation, that's a
sign things are headed in the right direction
Seigo and a partner started Make Play Live (MPL) to create "ethically correct
devices" that are hackable. A business ecosystem has also been built around it
to support the effort. The plan is to create a tablet called Vivaldi,
but there have been some problems along the way.
Seigo held up the tablet, noting that it was the second revision of the
hardware that was received from MPL's hardware partner. Using
that hardware, the company got 98% of the way there, he said, and were
demonstrating the device widely. It just needed a "little more polish"
before it was ready to ship. Then, the third revision of the hardware
The new hardware looked identical—on the outside—but was
"completely different" internally. MPL found out about the changes
after the fact, and was not able to provide input into the new
design. Because the volume of devices that MPL could promise to sell was
fairly low, the manufacturer had little interest in consulting or even
notifying the company about the changes.
The earlier revision had been running a modified Mer user space atop the Android kernel
distributed by the manufacturer, but that no longer works on the most
recent hardware revision. There is a "solution in the pipes" to that
problem, Seigo said, but that set Vivaldi back.
The device manufacturers don't really want to invest in Linux per
se, but want to focus on Android, which is a different thing.
In the Q&A session, Seigo further explained some of the problems that MPL
had run into. Unless it can promise a quarter of a million (or some other
six-digit number) of units, MPL won't be able to get any input into the
process. "Our order is a rounding error" on the total number of units the
device manufacturers are targeting. He certainly doesn't blame the
hardware companies as they are focused on their bottom line. It would be
great if MPL (or open devices in general) could rely on large companies to
take the baton and dangle that kind of volume in front of the
manufacturers, but that doesn't seem likely.
Part of the problem is that there is "little respect for the GPL" in Asia,
Seigo said. When you ask for the source to the kernel for a device, you
first get pointed at kernel.org. Once you make it clear that there is more
needed, you will get a tarball with "amazing stuff", some of which has
nothing to do with the device in question. Comparing that to what's
running on the device shows differences, so you have to ask: "Now can we
really get the source?". There is also often resistance from the hardware
to the whole idea
of getting the source as they think the company will go bankrupt if they
give it away. When setting out on this task, Seigo said that he had no
idea "how hard it would be to get GPL source" from the vendors.
The MPL partner network consists of nine companies so far who are
concentrating on various pieces of the problem, like human experience or
device integration. There is room for more hardware and software companies
in that network, he said. If some other company were to come out with a
an open device, he would see that as a success for the
project, even though it might be a competitor to Vivaldi.
The MPL philosophy is one of "human-centric experience" rather than the
"app-centric experience" offered by other mobile OS vendors (e.g. Google
and Apple). Vivaldi and other MPL devices are meant to be usable from
the outset and not require the purchase of a lot of apps. That
limits the "app store story" a bit, but it makes for a more compelling device.
When he puts a Vivaldi tablet into people's hands, they start immediately
talking about how they want to use it, Seigo said.
He noted that while tech pundits have written off the tablet market as a
two-horse race, they are not seeing the full picture because MPL devices
will not be competing in the same space. "If tech pundits were food
critics, they would be fired," he said. He likened the way pundits looked
at things to a food critic who said that French food is just great, so
"Italian food will never sell". Once devices start shipping, "we'll do just
fine", because MPL is not competing in the same space as iPad and Android.
When asked about where an interested person might be able to find paying
work on the MPL project, Seigo noted that some of the partners have been
employing people to work on it, as has his company, Coherent Theory LLC.
That model is not sustainable in the long term, but once devices start
shipping, there will be more money available for that kind of thing. There
are volunteers as well, of course; "not everything is done for
money", he said.
Enlisting aid from KDE developers and other interested people was one of the
themes of Seigo's talk. Much has been accomplished, but there is lots
still to be done. MPL needs "more people who care" to "join us and make
this a reality". He and others are committed to making open devices
available, with some help they can get there faster.
[ The author would like to thank KDE e.V. for travel assistance to Tallinn
for Akademy. ]
to post comments)