You are currently browsing the monthly archive for June 2009.

Andy V. O’Lay blogs this morning at what is possibly the most insightful blog around on mobile computing, VisionMobile. He is talking about the merger I referred to the other day between Wind River and Intel. Since O’Lay’s insights originate in such a different place from mine I though it would be useful to share the link. Go read his post, and then come back.

Andy says some very important things. Most notably:

“A reference design is not a half-baked breadboard; it is a scale 1:1 device, ready to ship.”

I am not certain I agree with him that TI is on the skids because of their lack of a serious software solution to their excellent lines of processors. But that is only because I don’t agree TI is on the skids at all.

“the key to hardware success is software.” …
“mobile software companies are instrumental in making silicon solutions pervasive, because they tick two major check boxes: reference design and support” …
“This is a visionary move. Hardware (HW) and software (SW) guys realising that they need each other to grow.”

I agree with it all except the last bit. This isn’t a visionary move. Sometimes it is forgotten in the heat of the market that hardware and software go hand in hand, always have, and always will, but it is never forgotten by the companies who make it happen. One reason I like embedded software is because I am constantly reminded of this unity. Making hardware and software engineers work together is certainly not new. IBM was doing this in the 1960s when the distinction between hardware and software barely existed {*}. Apple employs both hardware and software engineers, though they don’t produce their own chips (yet). Then again, there are also cases of companies spinning off one or the other, for business reasons of their own.

One thing is certain, though. This merger will continue to create opinions and counter-opinions from pundits everywhere. It remains to be seen whether it will change the landscape of mobile, embedded, Linux, open-source, or any combination of these.

{*} Historical note: the term “software” was coined in a mathematics journal in 1958 by Princeton mathematics professor John Tukey. The journal was the American Mathematical Monthly. Thanks to Ivars Peterson’s Mathtrek site for providing this valuable information.

Media coverage of embedded Linux has been thoroughly buzzing this week with Intel‘s acquisition of Wind River Systems, venerable RTOS experts and (recent) purveyors of embedded Linux. I have a strong interest in this particular event, partly because of my current position as developer advocate and Meld admin as well as technical writer at MontaVista, but also as an ex-Wind River employee (from the Sojourner days). In addition, I spent about five years writing documentation at Transmeta, a glorious little company that was thoroughly trounced by Intel—fantastic technical expertise, rather optimistic business sense. In other words, I have some investment in this announcement.

First, a little history. Intel has long been a player in the embedded space—I wrote documentation for GNU cross-development tools for the i960 as far back as 1992—but only as a side interest to their ever-booming x86 desktop and server business. Transmeta is widely acknowledged as having been the entity that forced Intel to address the low-power, low-heat market. The resulting competition ushered in entire new genres of devices that blurred the line between embedded systems and non-embedded, general-purpose computers. Intel’s investment in Moblin and the related introduction of Atom processors cements Intel’s commitment to following the thin-and-light market wherever it goes, leading wherever possible, and competing like a featherweight boxer with his girlfriend at ringside.

What this merger means for Intel is that their commitment to embedded (or at least thin-and-light) systems is tightly coupled with their commitment to Linux, the reasons for which are likely obvious to anyone reading this. Much will be made in the press and in the blogosphere in the coming weeks about Intel thumbing its nose at Microsoft by embracing Linux in this way. This may be true. WinCE succeeded in taking over market share from VxWorks and other embedded operating systems in the late 90s, which arguably propelled Wind River to redesign itself as an embedded Linux company. However, embedded Linux continues to sap market share from both WinCE and VxWorks and is slated to continue to do so on an accelerated scale, so the merger makes perfect sense for Intel. Those guys started out smart.

How does the merger benefit Wind River? That is a good question.

Wind River’s core business has always been real-time, with VxWorks and its ecosystem of tools. x86 is only a portion of that business, and though one might expect that it will become much more prominent now, it would be difficult for Intel to simply abandon a solid moneymaker, even one that centered around its competitors’ hardware. Realistically, the Atom is not the right tool for the job in many places where VxWorks shines, namely very hard real-time embedded—automotive, aerospace, and space systems in particular—noting especially that these are historically very conservative markets. I predict that, as Intel has suggested, they will continue to let Wind River operate in that environment as it always has, and will just take a paycheck, at least until those markets realize the possibilities that exist with Linux. (Automotive is already making strides in that direction.)

As for Linux, though, Wind River is known in the press for being an active purveyor of embedded Linux operating systems and tools, if not necessarily an innovator. This acquisition immediately distances Wind River Linux from non-Intel markets, namely ARM, Freescale, and Cavium. As Joerg Bertholdt (MontaVista’s VP Marketing) notes in a recent interview, current Wind River Linux customers who use non-Intel processors are already wondering what the future will hold.

However, Wind River started out smart as well. As their VxWorks business declines into hardcore niche markets and their Linux business—-well, “matures” rather than explodes as they might have hoped, they are intelligently seeking a solid rock to which they can anchor their core business values. I predict that VxWorks will continue to be available for a variety of architectures, though it may not continue to mature as it has. It seems easy to predict that Wind River Linux will mature in the direction of Intel hardware, though other hardware platforms may suffer, and that they will move much closer to Moblin and netbooks as a focus.

This is where things get interesting for MontaVista, who has always maintained a level playing field with regard to architectures. MontaVista Linux already supports dozens of architectures, Intel included, and will continue to do so under the new MontaVista Linux 6 Market-Specific Distribution paradigm. The MVL6 Integration Platform provides developers with an intense new method for creating, building, and maintaining their development environments, and DevRocket 6… (oops, shh). And Meld provides everyone a community to discuss it all.

In the end, the merger is fascinating and will be big news for both Wind River and Intel, and I will be very interested to see it play out, but I’m glad to be watching from the sidelines. MontaVista is well-poised to continue to enable embedded Linux developers to succeed, no matter what their choice of hardware, and as a Developer Advocate that’s what I care about.