The Boeing 787 Dreamliner remains just that: A dream transcending the material world by fiat—simply imagining something makes it so. The way that Web 2.0 has allegedly transmuted a boring world of clicks and page loads into a multimedia extravangza of Web-based applications. In this case, the 787 is the dream of bits defeating atoms. That's not how the fairy tale ends. Like many of the stories the Grimms drew on, it’s more about nightmares that well up from the underlying reality than puffy clouds and brows unwrinkled in restful sleep.

In the last few days, Boeing has admitted to a recently discovered serious design problem. The company says the problems uncovered can be fixed through a few patches. A Boeing supplier, not blamed for the problem, told the Seattle Times that the first test flight is now months away, nearly two years after it was originally scheduled. (Boeing hasn’t released a new timetable yet.)

Planes used to be constructed with a lot of a brute force, elbow grease, and intuition. My two boys make me a frequent visitor to the Museum of Flight—to which Boeing was a heavy contributor—where I can see how early planes were designed and tested, and the kinds of advances in process and materials that have occurred over many decades of humanity momentarily defeating gravity.

The 787 wasn’t the first plane to be designed entirely in computer software. The 777 has that honor. Its development included a lot of simulation that formerly would have required components to be built and beaten up. The 777 turned out well, has been quite successful, and seemed to be the right foundation for how Boeing would conceive, design, and simulate performance of planes from then on.

The 787 or Dreamliner added a couple of elements to the 777 program: Composite materials would be used for much of the plane, an idea well tested in aircraft but not large commercial ones. And the global supply chain used for assembly of the 777 would grow even bigger and more complex. This required modifying some old 747s—dubbed Dreamlifters— to carry fuselage components to Everett.

Modern communications would bind all this together. Any bit of data needed anywhere in the world could be instantly found. Even making calls among countries in 2004 was a damn sight easier than at the outset of the 777. The Internet, fast computers, and plastic goo would make the 787 fly better and cheaper.

Or maybe not.

When I first moved to Seattle in 1993, I met a Boeing employee at a party, and he explained that his job was to document via new information systems the tools used in building current planes. Or, rather, one step up. He was designing a system that would be used to store information about the tools used to build the planes.

The company was working so hard and fast at spitting out aircraft that specialized instruments needed to assemble planes were created with no method to track the design blueprints for those tools, or what those tools were used for. If an engineer retired, he or she might take all knowledge of the tool along.

I remember being floored by that, but I was naive about technology, Boeing, and bureaucracies. In the intervening time, I've learned how often companies eat their own projects, leaving no trace behind of how they succeeded (or failed) the previous time around. (This doesn't even scratch the surface of "siloing," in which different groups in a company—silos—defend their turf, and ruin collaboration.)

The 787 was supposed to be a leapfrog around that kind of problem. Computer systems had come a long way in the decade since the start of the 777 project. Need to make a new part or a tool to install it? In the realm of tools, for instance, hit a button and a person could get the plans to build that tool, or the plans would be set to an electronically controlled cutter. Or, shades of the future already made real in the present, a 3-D printer —technically, a rapid prototyping machine— could build or carve a complete tool or component of a tool. Potentially, that could happen anywhere in the world, too.

What went wrong? You can read the particulars on any given day in The Seattle Times, it seems. (That makes my boys happy, who love to see planes and trains in the paper, even when it's bad news.) But there were a lot of baffling mistakes to this technology geek's eye.

Despite a worldwide partnership, components and major sections arrived in Everett with large amounts of parts remaining to be assembled or mis-installed, without anyone at Boeing seemingly knowing beforehand. Apparently, the Internet didn't help them actually talk to one another. Or it led to complacency. It would seem no Boeing staff were on the ground when components were completed before they were packed up, or problems would have been known at that stage. Did meatspace get entirely replaced by cyberspace?

Design problems appear to have materialized at late stages, too, despite all the simulation that should have precisely prevented last-minute snafus. This last flaw seems nearly unconscionable, and while Boeing expresses optimism at an easy and relatively fast fix, it's hard to know how that's possible when so much prediction has gone wrong so far. You can read in great detail about the flaw elsewhere, but it stems from the wing design putting too much strain on what Boeing describes as a few square inches of fuselage here and there—enough strain, the company has implied, to distort or tear the composite material of the plane at those points. The solution sounds like a hack: like flower patches on classic jeans, hardly the image you want for the most advanced commercial plane ever built.

The simulation that saved money and time in the 777 process doesn't seem to have accurately predicted what's happening now. (Sure, there was a two-month strike, too, but that doesn't explain a full two years' delay.)

By turning the 787 into a digital project that would alchemically be transmuted from base bits into precious atoms, Boeing was also infected with the problems that dog all software projects. For a ripping examination of that topic - honestly, rather exciting - Scott Rosenberg's "Dreaming in Code" (2007, ISBN 9781400082469) is the software equivalent of Tracy Kidder's famous examination of hardware in Soul of the New Machine (1981).

Software projects can never actually be complete. There are always bugs that can't be crushed; feature creep (new features added at the last minute); technical problems that suddenly leap up and often crush an effort without warning. Rosenberg's book looks at a software pioneer's efforts via a non-profit to build new open-source calendar and contact information; the project completion is always about a year away. (A 1.0 finally came out after the founder phased out his financial and personal involvement.)

Boeing doesn't have that kind of flexibility. In aircraft design, you can't suddenly decide that carbon-fiber reinforced plastic was the wrong choice, and you'd rather go with aluminum. Or that seats need to be 19 inches wide instead of 17.25. That adds years to a project, and can result in hundreds of millions to billions of dollars in lost profit.

And, like software, it's impossible to predict what the outcome of the 787 project will be. Boeing assumes and I presume that problems will be solved, and planes will ship. But it's possible that unraveling one thread leads to another and another. Some of that happened already.

You can ship buggy code, but you can't launch a buggy plane.