Tuesday, August 14, 2007

Mobile LinuxWorld: Accessing the other Palm story

If Palm Inc. (née Palm Computing) is trying to find life after the PDA, its former alter ego (PalmSource) is trying to run to the front of the parade of adapting Linux to mobile platforms.

Of course, PalmSource was purchased in fall 2005 by Access of Japan, outbidding Palm (Inc.) and Motorola. This was only about six months after PalmSource started shifting towards an embedded Linux strategy with its money-saving strategy of buying China MobileSoft, a Nanjing-based Linux developer.

Today, the business cards in the booth say Access Systems Americas, but the address show the same Sunnyvale address of the former PalmSource.

“How did they hope to make money off of Linux?” is what I asked myself after the original sale, rebranding, and abandonment of ongoing Palm OS revenues. Actually, after visiting the booth (and doing some background research), I realized I already had the answer. What ASA (PalmSource) is doing is exactly what I would have recommended from my 7 years of studying “how do firms make money from open source” — and they didn’t have to pay me a dime.

ALP hierarchy diagramThe most obvious point was the Access Linux Platform (ALP) architecture diagram — which is a mixture of off-the-shelf open source (shown here in blue), technology developed by ASA released as open source (purple), and the proprietary code licensed by ASA (gray).

As I understand it, ASA’s open source contribution is mostly found in the “Hiker” project, which is intended to supply many of the key APIs missing from GTK/GNOME but necessary for writing real native Linux mobile applications. (Hiker is described in a PDF white paper and now has a downloadable tarball). The Hiker part was originally MPL (because it’s a better license) but OSS guru “Lefty” Schlessinger said it will be available Real Soon Now dual-licensed in LGPL(v2) to make the GPL-centric GNOME community happy.

As befitting all the ex-Apple people at ASA, its decision to give away some (but not all) technology parallels the early Apple Computer open source strategy, which I described in a 2003 research paper as “opening parts.” It also fits the general pattern I found in my subsequent larger study of open source business models, where firms sell some parts and give away others.

In addition to the opening parts, the other key point was that (as Bill Lee of ASA developer relations told me) “On this side [of the Pacific], we’re a services company.” Services (despite lower margins) are also common with open source companies. Services have historically been big in the embedded market because no two devices are identical. Unlike the PC industry (where commonality is sought to reduce costs), embedded device makers relish differences (and differentiation) as they seek to push the envelope. There are lots of reasons why hardware companies need help with embedded software, which is why there are companies like MontaVista.

In fact, (today) ASA doesn’t sell a Linux distribution, but assumes that most of its customers have an existing distribution from MontaVista or Wind River (which just won Palm Inc.’s business). This was probably the part that made the least sense: once upon a time PalmSource sold an entire operating system, so the dependency on MontaVista or Wind River doesn’t make sense in the long run. That seems easily fixed: four years ago, WindRiver was a proprietary embedded OS company and the ultimate anti-Linux — and then they got into Linux. So Access could buy (or sell to) an existing Linux distro, or build it from scratch.

ASA is evangelizing developers (using the process perfected by Apple in the 1980s) to support its new Access Linux Platform. (To run the installed base of 25,000 old Palm OS apps, the ALP has a Palm OS compatibility layer). I’m sorry I couldn’t make Tuesday’s developer event but they wanted money and besides I was out of town. In addition to courting ISVs, ASA has also won support as the preferred Linux provider to Orange, the large French mobile phone operator.

There was one more surprise as I riffled through the various ALP white papers, i.e. marketing brochures (in the dead tree version handed out at the booth but also available online.) To put it starkly, Access wants to be the Symbian of mobile Linux.

Specifically, their white paper “Mobile Linux - Going Native” contains arguments that could have been ripped from a Symbian brochure. For users of “smartphones” (a term practically invented by Symbian), it extols the importance of making full native applications (as with a PDA or PC) rather than using lightweight application frameworks like Java or BREW. One excerpt:

Applications that run in a “sandbox” are nice, but future users of smartphones will come to expect the performance and capabilities of applications designed to run in native mode.
So Symbian and Access are predicting a world in which native applications continue to matter, while Google and thus far Apple expect mobile devices to use device-independent rich internet apps (based on things like Ajax) to be the wave of the future. The jury is still out.

Technorati Tags: , , , , , ,


David "Lefty" Schlesinger said...

One minor nuance I'd like to correct, Joel: we don't really assume that customers will be using a kernel from MV or WR to the exclusion of other possibilities (e.g. one supplied directly from the silicon vendor, etc.) The ACCESS Linux Platform is intended to be "kernel agnostic": we can run our platform on top of any reasonable (in terms of general version, etc.) kernel. Internally, we use 2.6.18 currently, but we could move to a later revision pretty easily.

We can supply a kernel--the one we use for internal development--if that's what the customer (an ODM, for example) wanted, but you're correct in that we're not a "distribution vendor" in the sense that Red Hat or similar companies are...


David "Lefty" Schlesinger
Director, Open Source Technologies
ACCESS Co., Ltd.

Joel West said...

At LinuxWorld, nobody in the Access booth mentioned the idea of a BSP having a Linux kernel thrown in, but it certainly fits with being kernel agnostic.

I still believe that — commodity or not — Access will in the long run need to get its hands dirty by actively providing the complete solution, including the kernel. But that’s an inference based on past industry evolution — not based on talking to your customers and hearing directly what they value.