Linux Device Roundup

Four Linux device experts offer their opinions on the state of Linux devices and tell you about their must-have favorites.
What limitations or barriers need to be overcome for world domination by Linux in the devices space?

Kingman: Well, Linux has been the dominant device OS since 2003 or 2004, according to the market research I see. For it to continue, I see three hurdles to clear:

  1. Each vertical market may ultimately need to fight fragmentation by forming an industry group and/or support existing standardization efforts, like CELF, LiMo, the LF and so on. That way, the “value line” separating what you get for free vs. where you start differentiating your product can continue to rise, which in turn will attract new adopters. Linux is a collaboration, and I suspect that the better we work together, the bigger it will get.

  2. Within the limitations of the fast product development cycles embedded Linux developers have to deal with, it would be great if more of them would get involved in Linux kernel development. By “active role”, I mean tracking current kernel versions when feasible during development and submitting patches to the LKML. This saves Linux from being wholly shaped by enterprise server companies, but it also makes things easier when the time comes to port your stack forward to a new kernel version (if you ever plan to do that), especially if your patches are accepted.

  3. It would be great if things continue to get easier, financially, for the commercial embedded Linux OS and tool providers like MontaVista, Wind River and others. These companies contribute quite a bit to open-source projects, like Linux and Eclipse, as well as to standards bodies and industry groups, helping to ensure that embedded interests are well represented in key projects.

Lehrbaum: One area of concern is the lack of a truly dominant, mainstream, free graphical application toolkit for device applications based on Linux. Qt is popular, but there are licensing and royalty requirements for commercial device applications. This could result in design wins for Windows CE, which is offered at a fairly low royalty rate to OEMs, and which has excellent and inexpensive development tools.

Powers: If you look at the number of Linux-based devices in the market now—not a whole lot needs to be done. I think perhaps we need to market “Linux Inside” stickers, so people actually know there's a penguin under the hood.

If you look slightly outside the gadget or device arena, however, there are a few hurdles that are still sizable. I have a friend who works for Honeywell, designing hardware-testing solutions. Much of the firmware and software he develops for the hardware is still in MS-DOS. Many vendors are still embedding with DOS and providing only DOS drivers for their hardware. Since he's a friend, I've been hounding him for years to start switching over to Linux instead of building layers of DOS on top of Windows solutions, but in the end, he's held hostage by hardware vendors not opening up their specs enough to allow a programmer any access outside their proprietary DOS or Windows drivers. Because much of what he builds is mission-critical (and often could endanger lives), reverse-engineering isn't something he feels confident doing. It's very frustrating.

Weinberg: In real-world terms, “world domination” and 30%+ market share are almost synonymous. However, some ongoing barriers to adoption include:

  • The refusal by the kernel developer community to stabilize kernel APIs and driver architecture to help OEMs and OSVs “future-proof” device drivers. The mismatch between the practices of this important community and actual industry practices is a gap that neither side is ready to bridge.

  • The hazy definition of “embedded Linux” and “mobile Linux” that is presented to ISVs and other developers not currently familiar with Linux. Although Windows/CE/Mobile, Symbian and Java are actually rather fragmented of their own accords, they at least provide the appearance of unified APIs and SDKs. The situation for Linux is improving. See the emerging quasi-standardized APIs from LiMo, OHA and others.

  • The insistence of many embedded industry players, advocates and opponents of embedded Linux alike, in perpetuating concerns about GPL and other OSS licensing terms. Reciprocity need not be equated with “infection” and “contamination”.

Are there any “dud” devices out there that don't live up to their promise in your view?

Kingman: Any Linux device released without source code or a promise to provide it is a dud in my book. There tends to be more GPL license violations in the device world, I guess because people think that no one will notice or want to modify software that's “embedded” inside a device. But, it's pretty obvious which devices out there run Linux. Usually, you can tell from a glance at the spec sheet—let alone any of the more-technical telltale fingerprints.

Lehrbaum: The Nokia 770. I've had one since it came out and have updated to the latest released OS for it, but I have to say that its capabilities are quite disappointing—particularly in comparison to how well Apple's iPhone performs on a small touchscreen.

Powers: Well, my Eee PC's tiny keyboard annoys me, but my fat fingers shouldn't count as a strike against ASUS. If there's a potential dud, it would be with the saturation of Netbook solutions from multiple vendors, and multiple revisions from the same vendors. I'm not sure whether that means it's a dud or just the natural progression of a viral product idea, but I do worry that it will start to veer people away from the tiny Notebook concept.

Weinberg: It's best to reference the reviews on The biggest dud concept, however, is that Linux-based phones confer almost none of the virtues to those devices that the OS does to other devices.


James Gray is Products Editor for Linux Journal.