OSNews

As others have pointed out, this may be meaningless. The article points out that the sample was self selecting. Even if it was a randomized sample, it is difficult to tell what the uncertainties are.

But now to my main point.

If this shift is real, the question is why? Without knowing why, Linux developers cannot fix the problem, if there is even a problem to fix.

Fixable problems may fall in several categories. The two that I can think of off the top of my head: maybe the tried implementing an embedded device using Linux, but found that it was too difficult to do. If they can elaborate on those difficulties, it may be possible for external developers to find solutions. Of course, that is a huge thing to hope for since embedded systems cover an enormous range of devices. The other problem may be the reception of their customers. A case in point may be illustrated by those netbook return rates, which isn't really embedded but is approaching the realm of embedded systems. If the client finds it too unfamiliar or somehow unreliable, it may be possible to make the user interface more familiar or more underlying software more reliable.

Non-fixable problems may be suitability to task. If the market is shifting away from complex multi-function devices and towards less expensive single-function devices, then Linux is probably not suitable. Cost constrained applications may use a microprocessor that Linux won't even run on, or Linux may increase the demands on memory or the CPU to unrealistic levels. After all, some stuff can get away with an 8-bit microcontroller. Other applications may require a more powerful CPU, but you would save money if you could use 4 MB RAM instead of 8 MB of RAM.

Alas, surveys like this don't offer any insight into those sorts of issues.