Open source hardware: The problems and promise

Open source hardware projects have struggled to gain the mass audience that popular open source software projects have. This may not matter.

open source

Image: yuriz, Getty Images/iStockphoto

Is open source hardware a thing? No, I'm not asking whether open source hardware projects exist. Of course they do, whether high-profile projects like Open Compute (from Facebook) or lower-profile projects like the Homebrew D-STAR Radio (for amateur radio enthusiasts).

No, what I mean is, have any open source hardware projects taken off similar to Kubernetes, Apache Kafka, or Linux? And, perhaps just as interestingly, does an open source hardware project have to reach that level of prominence in order for it to be useful?

SEE: Inside the Raspberry Pi: The story of the $35 computer that changed the world (cover story PDF)

Open sourcing robots

The question arose during a conversation with Jason Huggins, founder and CEO of Tapster Robotics. As Huggins related, "I always wanted to play with robots professionally but had no idea how to get there." As a test automation engineer at Google and then as the co-founder of the Selenium open source project, a framework that enables browser automation and testing, Huggins was "on that path to get to his robotics future and just didn't realize it."

Tinkering, he built a button-pushing robot, one capable of replicating the human motion of pushing a touchscreen (and thereby playing Angry Birds back when that was a thing). In so doing, he caught the attention of Mercedes-Benz, which ordered 10 and put Huggins on his current path.

But what path is it, exactly?

Huggins open sourced Tapster's hardware and software, encouraging contributions (like this one from Pierre-Yves Lapersonne). There haven't been millions of contributions, however, or even hundreds. Does the fact that it's open source really matter?

On the one hand, the answer is clearly "Yes." Open source encourages a level of permeability into the mechanics of a project. This instills trust. As Huggins said of the early, unexpected success of his screen-tapping robot:

The robot wasn't like "I had 50,000 downloads and it was time to quit my job" but I started to realize that it wasn't a joke; not an art project. There were real people in quality control who make tangible, physical products, and they were having difficulty at the intersection between software and hardware.

By open sourcing Tapster, Huggins not only gave such people a product they could use to navigate this "intersection between software and hardware," but also the means to evaluate it at low cost, while minimizing their need to trust Huggins as its developer--they could inspect the hardware designs and associated code and see for themselves.

Is there a huge population of developers interested in screen-tapping robots? No. Need there be? I'm not sure. Early in open source's evolution, someone told me that open source could only succeed with a sufficiently large population of developers with interest and aptitude in a particular area. This, he explained, was one reason open source flourished in the operating system space (every developer needs an OS) and less so in ERP (no one writes ERP software unless they're paid to do so).

And yet, for Tapster, it could be enough to rally the few automobile manufacturers to collaborate on hardware designs specific to their needs. One wouldn't need millions of developers to make that collaboration worthwhile to the few.

Which brings us back to open source hardware, generally.

Does open source hardware matter?

Unlike open source software, open source hardware seems like it will always be at a bit of a disadvantage. Software, after all, can be free, requiring only an investment of one's time, but hardware will always have a cost. It takes up space. It must be shipped (or built using a 3D printer). 

SEE: The best alternatives to the Raspberry Pi (free PDF) (TechRepublic)

Looking around the industry for prominent examples of successful open source hardware projects, it's difficult to identify many or, really, any. At least, by the standards of a Kubernetes or Linux.

The most prominent example of an open source hardware project might well be Facebook's Open Compute. Released to much fanfare, Open Compute promised to open source Facebook's data center designs and enable every company to operate just a bit more like Facebook.

It hasn't worked.

As Jessie Frazelle has pointed out, Open Compute is "mostly deployed at Facebook and hyperscalers [and] definitely not very much outside of that." Why? As Frazelle continued, it's "too hard for others to adopt outside of hyperscalers [because] they'd need ODM [Original Device Manufacturer, like Qiku] relationships. Those take lots of time [to build and manage] and they'd need [a] hardware and firmware team." Companies already struggle to staff adequate software expertise. Adding in hardware expertise is just too hard for most.

And yet….

Mercedes-Benz is reportedly mulling over the idea of sharing its F1 designs. Such designs wouldn't be useful to 99% of the world's developer population, but for those few in the 1%...it could be a way for the industry to collaborate on base-level designs and save their engineering firepower for more advanced differentiation. The same could hold true for Tapster, as mentioned, as well as the few hyperscale companies taking advantage of Open Compute.

Open source hardware, in other words, doesn't need the mass appeal of open source software to be considered successful. All open source hardware needs is to connect comparatively few developers to collaborate on high-value projects, such that those developers needn't reinvent (and rebuild) the wheel.

Disclosure: I work for AWS but have no involvement (direct or indirect) with any teams that may be working on topics related to this post (if any). 

Also see