Paul Boddie's Free Software-related blog

Paul's activities and perspectives around Free Software

EOMA-68: The Return

It is hard to believe that almost two years have passed since I criticised the Ubuntu Edge crowd-funding campaign for being a distraction from true open hardware initiatives (becoming one which also failed to reach its funding target, but was presumably good advertising for Ubuntu’s mobile efforts for a short while). Since then, the custodians of Ubuntu have pressed on with their publicity stunts, the most recent of which involving limited initial availability of an Ubuntu-branded smartphone that may very well have been shipping without the corresponding source code for the GPL-licensed software being available, even though it is now claimed that this issue has been remedied. Given the problems with the same chipset vendor in other products, I personally cannot help feeling that the matter might need more investigation, but then again, I personally do not have time to chase up licence compliance in other people’s products, either.

Meanwhile, some genuine open hardware initiatives were mentioned in that critique of Ubuntu’s mobile strategy: GTA04 is the continuing effort to produce a smartphone that continues the legacy of the Openmoko Neo FreeRunner, whose experiences are now helping to produce the Neo900 evolution of the Nokia N900 smartphone; Novena is an open hardware laptop that was eventually successfully crowd-funded and is in the process of shipping to backers; OpenPandora is a handheld games console, the experiences from which have since been harnessed to initiate the DragonBox Pyra product with a very similar physical profile and target audience. There is a degree of collaboration and continuity within some of these projects, too: the instigator of the GTA04 project is assisting with the Neo900 and the Pyra, for example, partly because these projects use largely the same hardware platform. And, of course, GNU/Linux is the foundation of the software for all this hardware.

But in general, open hardware projects remain fairly isolated entities, perhaps only clustering into groups around particular chipsets or hardware platforms. And when it comes to developing a physical device, the amount of re-use and sharing between projects is perhaps less than we might have come to expect from software, particularly Free Software. Not that this has necessarily slowed the deluge of boards, devices, products and crowd-funding campaigns: everywhere you look, there’s a new Arduino variant or something claiming to be the next big thing in the realm of the “Internet of Things” (IoT), but after a while one gets the impression that it is the same thing being funded and sold, over and over again, with the audience probably not realising that it has all mostly been done before.

The Case for Modularity

Against this backdrop, there is one interesting and somewhat unusual initiative that I have only briefly mentioned before: the development of the EOMA-68 (Embedded Open Modular Architecture 68) standard along with products to demonstrate it. Unlike the average single-board computer or system-on-module board, EOMA-68 attempts to define a widely-used modular computing unit which is also a complete computing device, delegating input (keyboards, mice, storage) and output (displays) to other devices. It has often been repeated that today phones are just general-purpose computers that happen to be able to make calls, and the same can be said for a lot of consumer electronics equipment that traditionally were either much simpler devices or which only employed special-purpose computing units to perform their work: televisions are a reasonably illustrative example of this.

And of course, computers as we know them come in all shapes and sizes now: phones, media players, handhelds, tablets, netbooks, laptops, desktops, workstations, and so on. But most of these devices are not built to be upgraded when the core computing part of them becomes obsolete or, at the very least, less attractive than the computing features of newer devices, nor can the purchaser mix and match the computing part of one device with the potentially more attractive parts of another: one kind of smart television may have a much better screen but a poorer user interface that one would want to replace, for example. There are workarounds – some people use USB-based “plug computers” to give their televisions “smart” capabilities – but when you buy a device, you typically have to settle for the bundled software and computing hardware (even if the software might eventually be modifiable thanks to the role of the GPL, subject to constraints imposed by manufacturers that might prevent modification).

With a modular computing unit, the element of choice is obviously enhanced, but it also helps those developing open hardware. First of all, the interface to the computing unit is well-defined, meaning that the designers of a device need not be overly concerned with the way the general-purpose computing functionality is to be provided beyond the physical demands of that particular module and the facilities provided by it. Beyond such constraints, being able to rely on a tested functional element, designers can focus on the elements of their device that differentiate it from other devices without having to master the integration of their own components of interest with those required for the computing functionality in one “make or break” hardware design that might prove too demanding to get right first time (or even second or third time). Prototyping complicated circuit designs can quickly incur considerable costs, and eliminating complexity from what might be described as the “peripheral board” – the part providing the input and output capabilities and the character of a particular device – not only reduces the risk of getting things wrong, but it could make the production of that board cheaper, too. And that might open up device design to a broader group of participants.

As Nico Rikken explains, EOMA-68 promises to offer benefits for hardware designers, software developers and customers. Modularity does make sense if properly considered, which is perhaps why other modularity initiatives like Phonebloks have plenty of critics even though they share the same worthy objectives of reducing waste and avoiding device obsolescence: with vague statements about modularity and the hint of everything being interchangeable and interoperating with everything, one cannot help be skeptical about the potential complexity and interoperability problems that could result, not to mention the ergonomic issues that most people can easily relate to. By focusing on the general-purpose computing aspect of modularity, EOMA-68 addresses the most important part of the hardware for Free Software and delegates modularity elsewhere in the system to other initiatives that do not claim to do it all.

A Few False Starts

Unfortunately, not everything has gone precisely according to schedule with EOMA-68 so far. After originally surfacing as part of an initiative to make a competitive ARM-based netbook, the plan was to make computing modules and “engineering boards” on the way to delivering a complete product, and the progress of the first module can be followed on the Allwinner A10 news page on the Rhombus Tech Web site. From initial interest from various parties at the start of 2012, and through a considerable amount of activity, by May 2013, working A10 boards were demonstrated running Debian Wheezy. And a follow-up board employing the Allwinner A20 instead of the A10 was demonstrated running Debian at the end of October 2014 as part of a micro-desktop solution.

One might have thought that these devices would be more widely available by now, particularly as development began in 2012 on a tablet board to complement the computing modules, with apparently steady progress being made. Now, the development of this tablet was driven by the opportunity to collaborate with the Vivaldi tablet project, whose own product had been rendered unusable for Free Software usage by the usual product iteration performed behind the scenes by the contract manufacturer changing the components in use without notice (as is often experienced by those buying computers to run Free Software operating systems, only to discover that the wireless chipset, say, is no longer one that is supported by Free Software). With this increased collaboration with KDE-driven hardware initiatives (Improv and Vivaldi), efforts seemingly became directed towards satisfying potential customers within the framework of those other initiatives, so that to acquire the micro-engineering board one would seek to purchase an Improv board instead, and to obtain a complete tablet product one would place an advance order for the Vivaldi tablet instead of anything previously under development.

Somehow during 2014, the collaboration between the participants in this broader initiative appears to have broken down, with there undoubtedly being different perspectives on the sequence of events that led to the cancellation of Improv and Vivaldi. Trawling the mailing list archives gives more detail but not much more clarity, and it can perhaps only be said that mistakes may have been made and that everybody learned new things about certain aspects of doing business with other people. The effect, especially in light of the deluge of new and shiny products for casual observers to purchase instead of engaging in this community, and with many people presumably being told that their Vivaldi tablet would not be shipping after all, probably meant that many people lost interest and, indeed, hope that there would be anything worth holding out for.

The Show Goes On

One might have thought that such a setback would have brought about the end of the initiative, but its instigator shows no sign of quitting, probably because genuine hardware has been made, and other opportunities and collaborations have been created on the way. Initially, the focus was on an ARM-based netbook or tablet that would run Free Software without the vendor neglecting to provide the complete corresponding source for things like the Linux kernel and bootloader required to operate the device. This requirement for licence compliance has not disappeared or diminished, with continuing scrutiny placed on vendors to make sure that they are not just throwing binaries over the wall.

But as experience was gained in evaluating suitable CPUs, it was not only ARM CPUs that were found to have the necessary support characteristics for software freedom as well as for low power consumption. The Ingenic jz4775, a sibling of the rather less capable jz4720 used by the Ben NanoNote, uses the MIPS architecture and may well be fully supported by the mainline Linux kernel in the near future; the ICubeCorp IC1T is a more exotic CPU that can be supported by Free Software toolchains and should be able to run the Linux kernel in addition to Android. Alongside these, the A20 remains the most suitable of the options from Allwinner, whose products have always been competitively priced (which has also been a consideration), but there are other ARM derivatives that would be more interesting from a vendor cooperation perspective, notably the TI AM389x series of CPUs.

Meanwhile, after years of questions about whether a crowd-funding campaign would be started to attract customers and to get the different pieces of hardware built in quantity, plans for such a campaign are now underway. While initial calls for a campaign may have been premature, I now think that the time is right: people have been using the hardware already produced for some time, and considerable experience has been amassed along the way up to this point; the risks should be substantially lower than quite a few other crowd-funding campaigns that seem to be approved and funded these days. Not that anyone should seek to conceal the nature of crowd-funding and the in-built element of risk associated with such campaigns, of course: it is not the same as buying a product from a store.

Nevertheless, I would be very interested to see this hardware being made, and I am even on record as having said so. Part of this is selfishness: I could do with some newer, quieter, less power-consuming hardware. But I also think that a choice of different computing modules, supporting Free Software operating systems out of the box, with some of them candidates for FSF endorsement, and offering a diversity of architectures, would be beneficial to a sustainable computing infrastructure in the longer term. If you also think so, maybe you should follow the progress of EOMA-68 over the coming weeks and months, too.