Tech —

Whither Windows? OLPC 2 likely to use ARM, not x86

OLPC founder Nicholas Negroponte has revealed that the project could switch to …

The One Laptop Per Child (OLPC) project, which seeks to produce low-cost laptops for education, revealed on Thursday that it might drop the aging AMD Geode processor it has been using and replace it with ARM chips in its next-generation mobile computing device.

OLPC, which emerged from the MIT media lab and was founded by Nicholas Negroponte, has been plagued by an endless stream of technical problems and logistical blunders. The cost of the hardware, which was originally targeted at an ambitious $100, has climbed by 88 percent. The software platform strategy has totally disintegrated, leading to significant friction within the organization. Key figures have left in frustration and have vocally criticized Negroponte for taking actions that conflict with OLPC's stated goals.

In January, OLPC downsized half of its staff and largely discontinued its involvement in the development of the Sugar software framework—an education-oriented software environment built with open source technologies that is designed to run on the Linux operating system. Negroponte says that the cuts were needed because the organization's large corporate sponsors are closing their wallets in the current economic climate. In addition, the second Give 1 Get 1 program, a major source of funding for the project, fell flat and generated only a fraction of the revenue that was produced by the problematic first attempt.

OLPC could simply not afford to continue without major cuts and a serious change in strategy. Negroponte attempted to reorganize the group and shifted the focus away from large-scale distribution. The new approach is to produce hardware designs that can be manufactured and brought to market by commercial partners. This approach is similar to the one that Intel has pursued with its Classmate PC, which has already shipped to students in greater number than the OLPC XO laptop.

OLPC says that they intend to use ARM processors for their next-generation hardware because the ARM chips will use less power and could potentially deliver better performance than the aging AMD Geode that runs the current OLPC model. This seems like a reasonable move in principle, but it makes little sense when viewed in the context of the organization's software strategy—though it's hard to tell what the software strategy is even supposed to be anymore.

When OLPC began building closer ties with Microsoft and publicly criticizing Linux in press interviews, concerns were expressed by OLPC members that Negroponte might be preparing to abandon Sugar. He vigorously denied it and claimed that OLPC was planning to significantly increase its commitment to Sugar. Only months later, he reversed this position when he ended OLPC's involvement in Sugar development and turned it over to the community. Ongoing Sugar development is handled by Sugar Labs, which is run by former OLPC president Walter Bender.

The problem with moving to ARM is that Windows XP, which OLPC has increasingly been warming up to, won't run on it at all. Even the existing, Fedora-based Sugar platform would probably require some tweaking. Fedora's ARM port is promising, but still somewhat experimental. I suspect that the Fedora ARM port is mature enough that it could be adapted to meet OLPC's needs within a reasonable timeframe (an alternate approach would be to make Sugar and OLPC's software customizations work well on Canonical's officially supported ARM port of Ubuntu), but that would still require at least some software engineering. It's unclear if Sugar Labs is in a position to handle that kind of development at this stage.

Negroponte seems convinced that a Windows port for ARM would be advantageous to the project. Microsoft might have an incentive to port XP to ARM in order to prevent Linux from gaining ground on the inevitable commercial ARM-based netbooks, but it seems more likely that Microsoft would rather invest in moving forward with Windows Mobile for that market. Microsoft seems to have a clearly defined roadmap for Windows Mobile 6.5 and 7, but Negroponte isn't crazy about the idea of using Windows Mobile for his device.

"Like many, we are urging Microsoft to make Windows—not Windows Mobile—available on the Arm. This is a complex question for them," Negroponte told IDG.

Some pundits are speculating that Google's Linux-based Android platform could potentially be an option for an ARM-based OLPC device. Although Android could probably be made to run on the hardware, it really doesn't seem like a feasible choice. As Google's engineers have pointed out, Android uses the Linux kernel but is fundamentally not a Linux platform.

The Android stack doesn't support Xorg and is not capable of running conventional desktop Linux software. It's absolutely not equipped to accommodate Sugar (which is built with Python and GTK+) and there is no conceivable glide path for making such a port.

The hardware plans—a device with a dual-touchscreen interface and an extremely low-power ARM chip—are intriguing. I think that a device of that nature could have a lot of practical value if it's designed well. However, the general lack of direction at OLPC, the lack of staffing, the lack of resources, and the lack of a successful track record don't really bode well. The fact that Negroponte is publicly begging Microsoft for a Windows ARM port also adds to my skepticism.

The entire effort appears to have been in a steady downward spiral during the past year, and the potential for recovery looks slim. The latest revelation—an unplanned architecture change without any clearly articulated software strategy—seems to further erode the odds of success.

Channel Ars Technica