On March 5, 2001 CNET ran an ``other hand'' story about embedded Linux. The headline says ``Maxtor picks Windows, dumps open source''. The subhead explains, ``Maxtor has switched from the open-source FreeBSD version of UNIX to a special-purpose version of Windows 2000 for a new storage system it's introducing Monday.''

We're not going to make a distinction between embedded Linux and embedded BSD here. The fact that both are open source and well understood by UNIX programmers doing embedded work is what matters. For single-purpose embedded applications like Maxtor's--a network attached storage (NAS) device--the OS usually isn't a ``platform'' built to run a bunch of other applications. It's fundamentally a device driver that needs to live in the network-attached world, which is coming to include pretty much everything. In other words, the OS is a TCP/IP stack on a device that does very few things but does them very well. The OS can be anything the designer chooses. In this case, Maxtor chose Windows.

``The news is sour for advocates of open-source technology, who argue that their cooperative development model is more responsive to customer needs than the proprietary philosophy that underlies Microsoft products'', the piece continues.

Is this true? I work in the midst of hundreds--heck, thousands--of open-source advocates, and I can't think of a single one for whom this news would be ``sour''. I also failed to detect a single sour response a month earlier when Dell said it would make a NAS product that would use ``a version of Windows 2000 Advanced Server with many unnecessary components removed to increase performance''.

Embedded devices aren't PCs. Sure, politics may influence engineering decisions, but so will inertia and relationships with suppliers and partners. However, most decisions involve the highly arcane nature of the device itself. The engineers at Maxtor and Dell know what they're doing. Steve Wilkins, the product marketing director at Maxtor, told CNET that its new MaxAttach 4000 had ``required software features'' that favored choosing Windows. He also said that Microsoft ``adjusted'' its licensing terms for Maxtor, including a waiver of client-access fees. ``That's the first time Microsoft has done this'', Wilkins said.

There's your news.

Microsoft can't get anywhere in the embedded world unless they make their OSes more embeddable. Componentizing each OS and adjusting its license is only the beginning. To be fully useful, the source code needs to be open. That's exactly where Microsoft is headed.

Consider these two data points: 1) At COMDEX Fall 1999, Microsoft CEO Steve Ballmer said this in response to a question about the Linux ``threat'': ``We're going to think about the issue, and I don't think the advantage to Linux is that it's free. The issue is their flexibility in opening source code and how do you do that in an NT environment? And we're considering that.'' 2) In February 2001, Microsoft disclosed the ``Windows Embedded Strategic Silicon Alliance'' with a long list of major chip makers (pretty much everybody but Motorola), with the purpose of getting them to design for a new version of Windows CE named Talisker (apparently named after a Scottish distiller of single-malt whiskeys).

Only a few days earlier, another Microsoft executive revealed that the source code to Windows (not just CE) had already been made available to ``hundreds'' of OEM customers. Significantly, he said this while hanging around the LinuxWorld trade show in New York. (What would he be doing there? It's not exactly next door to Redmond.)

Meanwhile, Sharp said it will come out with a Linux-based PDA in October 2001 and hopes to sell a million units worldwide by March of 2002, which is about equal to the projections for Japan alone. They want developers to program in Java and hope to see 10,000 Java programs running on the device by October 2002.

Context: Sharp's handheld devices have run the company's own Zaurus OS since 1993, and it has been marketing PDAs in the US that run on Windows CE. The devices are not a big hit. Sharp went with Linux in part because it didn't want to be ``restricted'' by Palm or Microsoft, which would force them into price competition with others using the same OS. In other words, they were free to add more of their own value on Linux.

At this point it would be easy for us to haul out the war and sports metaphors and tell the usual story about how Microsoft and Linux are beating the crap out of each other for ``shares'' of the NAS and PDA ``markets''--or for the embedded ``space'' as a whole. The problem with those metaphors is that the entire embedded space looks more like the Big Bang than a battlefield or a sports arena. This doesn't just mean that there's room here for everybody. It means that everybody here is making room. Lots of it.

We believe (and we're hardly alone) that more designs will be embedding Linux than Windows, but that won't make Microsoft a failure in the new embedded universe. Linux is well-proven building material. Open source is a well-proven building method. Microsoft is learning from both, and it has little if any leverage for locking in customers. That's good for everybody.