cts | Volt a Day | Building Cars is not Publishing Software December 03, 2010

Really, one hopes that GM has looked carefully at the example of Eclipse Aviation. This is the story of a software business expert (Vern Raburn) who believed that he could copy his success in the software industry in the aviation industry. It turned out that daily builds of an operating system were very different from the daily test flights of a small jet-powered aircraft. His belief that he could bring a different perspective to a tiny market and beat out companies like Cessna, Beechcraft and Bombardier, all with decades of experience in the field, was blinding. It was a tragic effort to watch unfold.

There have been a lot of stories about the ten million lines of code in the Volt (you can Google the phrase to find some). That's a huge code base to manage (famously more than is in the F-35 advanced jet fighter). The moment there is a new Volt model, there will be hardware and software fragmentation, while the codebase has to fork to support more than one version of the vehicle platform. That's a nightmare of software development. Even without a car development alongside of it, it would be a major software development effort.

One of the elements of the EV1 story that struck me was when they were discussing some of the initial electronic components and the engineers at GM explained to one of the outside innovators that for GM a component that was demonstrated to work wasn't that important. It wasn't even good enough if it was demonstrated to work in the torturous range of environments that GM set up for their vehicles. They had to be able to extrapolate from the build and design of the component that it could last through ten years and a hundred thousand miles of driving. Essentially, that it was a rugged lifetime part. Otherwise there was a bean-counter sort of cost accounting which had to come into play: how many of these were going to fail, what would they cost to replace under warranty, and that affected the total cost of the vehicle.

So, although they are faced with one part of the product which is governed by (now) well-known rules of software development, GM will want to be able to ship the car out the door of the plant as if it is complete, and immutable, that the effort on that particular item can be contained and backstopped.

That would be a mistake.

They should commit to a software update schedule. One of the brilliant marketing moves of Apple is to hype each release of its operating system. At first glance this might seem like a strange decision, because Apple is a hardware company. (Initially a computer hardware company, and now a consumer electronics hardware company.) Apple doesn't make money selling music on iTunes, renting you movies on your iPad, and certainly not by selling the $99 operating system upgrades. But they have high-profile events to tell their customers what new features to expect in June, or January, or whenever the update is due, because they know that the moment people boot up the software they are aware of missing features and bugs. It is Apple’s way of reassuring the purchaser that they are not locked into something obsolete, that the software can improve even while they own the same hardware.

So GM should announce that they will release new software for the Volt every six months for the first three years. And for every year after. Somewhere in the Volt screen there should be an easy-to-find software version number, so owners are happy to know they have 1.1 or 2.6.

I admit that this is not a small commitment. It is a vital one. It is an admission by a giant car maker that it is producing something other than another piece of rolling stock. Much the way the offer a power train warranty, they should offer a software warranty, but the real way to do that is to promise continuing development. And part of that promise is smart protection against fragmentation (see Google's Android operating system for an example of how hard this is to avoid) or stagnation due to hardware incompatibility.

I have yet to write up a comprehensive review of my time with the Volt. I hope to do that in the coming week. I am not putting as many miles on the car as my fellow CAB members, but I believe I have a handle on what is good and what needs work. The software has the most glaring problems including a few "show stoppers" that I would not allow it to hit the road with. But I'm picky.