21. Push technology. In 1992, PointCast had a clever idea: Why not make it possible to view stock quotes, headlines, and other information in real time, without browsing the Web? Instead, the PointCast client would "push" the information direct to the desktop, all day long.

The idea spawned a horde of imitators. Unfortunately, no one foresaw the strain that all that pushing would place on the limited Internet connections of the time. Network managers banned the client, and modem-based home users balked at the ads being pushed to them along with their sports scores.

News once offered US$450 million for PointCast. Two years later, the push craze had evaporated, and it sold for a paltry US$10 million.

20. Copland. Some fumbles can be recovered. And it's true; today, Mac OS X is an impressive operating system. But imagine how much further Apple could have gone if it had delivered its next-gen OS when it originally intended to, back in 1995.

Copland was meant to be the modern successor to the original Mac OS, but years of political infighting had hobbled Apple's development department. For all its superior engineering talent, it became clear that it was impossible for Apple to produce a modernized Mac OS on its own. Instead, it would buy Steve Jobs' NeXT OS and use that as the basis for the Mac OS X that ultimately shipped in 1999 -- ironic, considering that Jobs had left Apple over political infighting a decade earlier.

19. Gnu Hurd. When Richard Stallman launched the Gnu project in 1983, his goal was to build the world's first completely free operating system: kernel, tools, utilities, applications, documentation, and all. Good thing he didn't start from the bottom up.

Almost 25 years later, there is still no Gnu kernel. The Hurd, as the proposed kernel is known, should have been the Free Software movement's crowning achievement. Instead it's become the poster child for collaborative software development gone wrong, topping the lists of vaporware year after year. And it's a shame -- because wouldn't it be great if there was a free OS kernel for everyone to use?

18. Oracle Raw Iron. What's the best OS for your database server? Should you run it on Windows? Linux? AIX? Something else? Back in 1998, Oracle's answer was none of the above! Instead, Larry Ellison promised an "appliance" version of Oracle 8i, called Raw Iron, that ran atop the bare server hardware. No longer would Oracle customers need to worry about a separate support contract with an OS vendor: Oracle would handle the whole show.

Behind the scenes, prototype Raw Iron boxes ran a custom version of Sun Solaris, but it didn't matter. Customers had seen through Larry's hand-waving, anyway. When nobody bit, the project was quietly shelved -- just a few years before the market for network appliances took off.

17. B-to-b e-commerce. As the dot-com craze waned in the early 2000s, venture capitalists clung to a last-ditch idea: If all those startup e-commerce companies weren't striking gold with the consumer public, maybe they could ply their wares to other, more established companies instead? They called it b-to-b e-commerce, and a generation of would-be digital disintermediators was born.

The problem was that few of their potential customers were interested in cutting out the middlemen -- not if it meant trading them for an unproven online startup with a tiny sales force and no real experience in inventory management. In the end, though, the b-to-b players did deliver some excellent deals -- when their assets were offered up at auction.