Nicholas? [LWN.net]

Nicholas?

Posted Jan 8, 2009 3:06 UTC (Thu) by mjg59 (subscriber, #23239)
In reply to: Nicholas? by BrucePerens
Parent article: Changes at OLPC

The brought up a new hardware platform, significantly later than expected and still buggy in fundamental ways (the mesh network still doesn't work properly, resulting in many of the advertised features being impossible - dcon has little quirks like not firing a vblank interrupt during the vblank) and they provided a software platform that's been rearchitected more times than I've shaved in the past 12 months. This isn't a criticism of the ability or dedication of the developers, just of the process that led to the failure to achieve the (admittedly aggressive) goals. From the outside, it looks awfully like better management would have made a big difference.

Don't get me wrong, I think OLPC has been a great achievement and the extent of its real world success has impressed me. But it's far too easy to just buy into the "Everyone who associates with Microsoft is doomed to failure as a result" story, and doing so doesn't actually help us avoid the real problem in future. A strong volunteer community can accomplish many things, but it won't improve the quality of your management.

to post comments

Nicholas?

Posted Jan 8, 2009 7:33 UTC (Thu) by dlang (guest, #313) [Link] (5 responses)

seperate out the software problems from the hardware problems, especially when you are accusing them of failing on the hardware side.

the mesh network issues are almost entirely software problems. this was made worse by the tunnel vision of kids working in the wilderness and forgetting that they also needed to work in a super-high-density school setting.

I'll admit I hadn't heard of the dcon problem, (the touchpad issues seem far more significant)

now that other distros are starting to be available to run on the XO I expect the software side to shape up quickly (nothing like competition to spur progress)

I agree that the microsoft dealings were not a major cause of problems (other than PR problems, which are noticable), NIH syndrome and lack of compatibility had far more to do with their problems.

but they aren't dead yet. they shipped a half million laptops last year (not counting the 2008 G1G1 program) and the factory will still be building more.

Nicholas?

Posted Jan 8, 2009 12:40 UTC (Thu) by dcbw (guest, #50562) [Link] (4 responses)

From somebody who worked on it, a huge problem with management was moving goalposts and featuritis, because if they didn't have feature X right away they weren't going to be able to close the deal with country Y, thus every single one of our goal-directed monthly snapshots was blown by 2 weeks of work on a new hot-button feature that usually proved pointless or a lot less important.

Second, because management kept talking to a number of quite different countries with quite different requirements, they weren't able to distill the *actual* software goals down to something managable. Had they decided to only target kids without a roofed school in the bush (or only kids in higher-density schools with WiFi APs), it would have been a lot easier to nail. Instead, to sell the laptop, management had to have *both* of those. And that's only one example of this behavior and thinking.

Nicholas?

Posted Jan 8, 2009 20:10 UTC (Thu) by branden (guest, #7029) [Link] (3 responses)

This sounds really familiar. Let me tell you a similar story.

I can think of a commercial Debian derivative that was hideously late in launching because of factors like this. "[A] huge problem with management was moving goalposts and featuritis, because if they didn't have feature X right away they weren't going to be able to close the deal with [hypothetical target market]*."

We went back and forth repeatedly, suffering identity crises over whether we were targeting "the enterprise" (because big, high-margin contracts are where the money is), or "grandma" (because high-volume, low-margin retail sales are where the money is). We slipped horrendously. I don't remember now but I'm pretty sure it was over 6 months.

Ultimately, while the product did finally ship and actually got good reviews, we captured neither market, laid off two-thirds of the staff, and fundamentally changed the business model of the company.

* While our founder was by no means an unknown figure, when one's brother formerly oversaw right-wing death squads in Honduras, one likely commands a degree of attention from, and shares a common language with, officials in third-world countries who can get things done.

Nicholas?

Posted Jan 8, 2009 20:40 UTC (Thu) by BrucePerens (guest, #2510) [Link] (1 responses)

You didn't mention the part about investing a ton of money in distributed filesystem development which became meaningless as disks got bigger.

Nicholas?

Posted Jan 11, 2009 5:24 UTC (Sun) by branden (guest, #7029) [Link]

No, mainly because I wasn't on that team, but the distro one.

Oddly enough, something useful did come out of that work. The "nullfs" kernel module later became part of a customer's proprietary backup solution.

When I look back on my years there, the failures I see were primarily not technological in nature. I don't know if that's good fortune, or an experience most software developers have. I've seen some ugly, ugly code, more of it proprietary than free.

Even dpkg is better described as "self-consciously eccentric" than "ugly".

Nicholas?

Posted Jan 9, 2009 13:34 UTC (Fri) by csigler (subscriber, #1224) [Link]

"... moving goalposts and featuritis...."

FWIW, in the business sector (and esp. IT) in the U.S., this frequently encountered problem is jokingly (and dyslexically :) referred to as:

"Feeping Creaturitis"