2008-04-23 15:12 | fche blog tech olpc conundrum
The current crisis in the OLPC project is painful to watch. Here are some of the complications, as I see them.
The XO computer’s hardware is widely lauded as revolutionary. Indeed, its particular combination of features (small size, robust construction, relatively cheap, built-in accessories, hybrid screen) is unusual and attractive. But:
- Initial power management plans (tens of hours of use time) is many times off reality (2-4). Suspend/resume takes seconds, so can’t be as aggressive as necessary.
- The proprietary network hardware (separate processor doing mesh networking) is turning out to be less useful than imagined, considering the nature of the software (heavily mDNS based, requiring an always-on OS to avoid having nodes disappear).
- There are silly glitches like a keyboard stuck keys, flimsy buttons, battery charging hardware/software problems.
The XO software is also a mixed bag. It has some interesting code, but:
- The web browser is a hobbled subset of mozilla. Its support for real flash is notoriously poor – no youtube, no kid flash games.
- The “sugar” gui is bogglingly slow, and does not readily allow preexisting linux applications to run.
- The software stack, down the OS, seemed to start with a full desktop distribution, being very gradually whittled down to eliminate whatever the gui & the few apps won’t need. It probably would have been better to start with the smallest possible linux package from 10 years ago, and adding the bare essentials. (Big desktop hardware of that era was roughly as powerful as the XO is today.)
- The XO software/build system is mostly open, but it is still difficult for someone to rebuild and install the whole OS. The XO is not self-hosting (too small/slow), so that end-users cannot even start such a rebuild.
The current crisis seems to be about the possible severing of the XO software into two bits, with the bottom (the bulk of the OS) being deemed expendable and replaceable with Windows; the top (“sugar’) being deemed worth keeping. So what is a fan of the XO do now?
- Those awestruck by the hardware have had to face the reality that the current batch is in many ways “beta quality”. Other hardware vendors — big companies — are coming in that are nearly as good, and can run linux (and thus most of the XO software).
- Those awestruck by the “sugar” shell have had to face the reality that this part of the software, such as it is, can probably be ported to whatever underlying OS is required — even windows. Several other novel ideas like the “bitfrost” security system may never get fully built — and, perhaps, just as well.
That seems to mean that most novel parts of the OLPC XO project are individually replaceable. And yet, to this curmudgeonly admirer of OLPC, it still somehow seems worthwhile to press ahead. Maybe it’s all up to outside admirers/developers whose assistance appears to be essential — but need to remain un-dis-illusioned.