Let's talk about the software business. We'll start with Microsoft. Hey, it isn't 2010 yet. If we're going to talk about the software business in 2001, we have to address the incumbent Kahuna. It isn't easy in any case, and this past February, Jim Allchin, who runs Microsoft's platform products group, made it a lot harder. In an interview with Bloomberg, he called Linux “an intellectual property destroyer” and open source a “threat” to “the American way”, among other wacky, paranoid things.

This draws attention away from much slower and less newsy changes happening at Microsoft for years and away from signs that the company is noticing its markets are fully networked and, therefore, behaving progressively less like “targets” for marketing messages and disposable consumer electronics—and more like bazaars.

A bazaar is a real market: a chaotic and noisy public place where all the reciprocal abstractions of economics—supply and demand, production and consumption—are a handshake apart. In real markets, clues from customers have a lot more impact on vendors than any vendor's “marketing strategy” has on “consumers”. Since marketing likes to be “strategic” and doesn't like to touch customers (that's sales' job), it becomes progressively more useless in the bazaar environment. “Consumer marketing” becomes an antique conceit, and something else takes its place; something that involves participation in markets at a much deeper, more organic level.

Something similar started to happen at Microsoft in the summer of 1996, when the company did an unusually clueful thing: they created unmoderated Usenet newsgroups for every one of their major products. Suddenly there were conversational zones where customers, employees and everybody else could say what they pleased. Some people posted suspicions that Microsoft had created a habitat for astroturfing or otherwise spinning untruths about its products (something the company has since been accused of doing in other venues). But employees jumped in and encouraged honest exchanges. They wanted to know as much as possible about how their products worked in the real world. A Microsoft executive at the time told me this was “Marketing 101” and expressed surprise that competitors weren't doing the same kind of thing.

However, Marketing 101 in those days was still mostly an academic exercise. Theodore Levitt taught that the purpose of marketing was “to satisfy the customer no matter what”. For all the clues it got from Usenet and tech support calls, Microsoft continued to make crash-prone operating systems. With good reason—they sold by the zillions. Money spoke and the market approved.

Starting in February 1998 though, the market started to speak in forms other than money. This was when Salon.com published the results of a “Haiku Error Message” contest. Within minutes of publication, the winning haikus were being mailed and posted all over the world. Although it didn't win first prize, this one by Peter Rothman unleashed one of the most viral memes in the history of the Net:

Thanks to the Net, everybody's screams became one huge joke. Windows became a target of mass mockery. Worse, it became a mirror in which it mocked itself. Look up “blue screen of death” in a search engine and near the top of the list you'll find a page (http://www.daimyo.org/bsod/) where Matt Michie (who works at Linux.com) keeps a rogues' gallery of public BSODs. He has shots of giant Las Vegas electronic billboards, ATMs, promotional marquees, airport arrival/departure displays and casino game machines. Naturally many of the same pictures appear on lots of other sites. Memes are like that. And the bigger the bazaar, the faster word spreads through it.

At some point, the din of clues overwhelms everything else. Clearly this happened last summer when Jim Allchin went on sabbatical with a bunch of friends in a boat. Recently, he told me he learned two things on the trip that amazed him: 1) how many different kinds of things people want to do with PCs; and 2) how much trouble they have making their PCs work. He called the experience “humbling” and said he was “more determined than ever to do something about it.” Note the word ever. This means he's been trying to do something about it for a long time, which would only make sense; whatever else anybody calls him, Jim's still an old UNIX guy.

Now we see Microsoft ads promoting “five nines” (99.999%) reliability for Windows 2000 Professional. The claim may be debatable but not the fact that reliability is a concern, as we've seen Microsoft suggest previously. Why does reliability matter now?

I think it's because the bazaar is losing at least some of its appetite for PCs marketed as consumer electronics. Starting in 2000, more customers began to realize last year's PC was still fine. So were Office 98, Quicken 7, Photoshop 5.5 and lots of other software products—perfectly adequate (if not always perfectly functional).

Then, there was the repositioning of the operating system. It used to be a pretty product. Now it's just infrastructure. For more and more people, the platform that came to matter was the Net, not the PC's OS.

And then there was the undeniable success of the Net's self-ubiquitizing infrastructure, which owed far more to free and open-source software development methods (and developers) than to anything coming from software vendors whose first urge is to lock in customers. Linux is clearly part of that. There is no way Microsoft could ignore the success of Linux or the growing importance of constantly building infrastructure.

Ubiquitous infrastructural building materials are by nature commodities. You might be able to make money selling them, but the margins are elsewhere. Still, as every business comes to depend increasingly on commoditized infrastructure, including the Net and Linux, there is a strategic imperative to contribute to that infrastructure. If you're not involved in building infrastructure, your know-how falls behind, to say nothing of innovation.

Microsoft is starting to get a little bit strategic about this stuff. That's why they worked together with other companies on the SOAP and XML-RPC protocols. This is a drop in the bucket compared to what Linux community natives like Red Hat, VA Linux, Caldera and others have been doing for years. But it's the same strategy.

So what's the strategy for Linux companies, which have been in the business of building commodity infrastructure all along? We know they can sell high-margin hardware and services. But can they sell high-margin software?

Yes. They can do this by keeping the high-margin source code closed and giving away as much open-source code as they possibly can. This is what Caldera is doing with Volution. It's what Borland is doing with Kylix. It's what Turbolinux is doing with TurboCluster. There is a huge open-source side to each of those companies' product strategies—one that involves giving away a lot of stuff. But Volution, Kylix and TurboCluster are all closed-source products. None of those companies go out of their way to point out that fact, but there it is.

Traditional closed-source companies like IBM, Sun, Borland and even Microsoft are starting to adopt open-source strategies (of very different sorts, at very different speeds). Do traditional open-source companies need to move in a closed-source direction if they want to sell high-margin products to enterprise customers? So far, their answer is yes. And they all know the arguments for open source.

So, until some company risks selling open-source software to enterprises at a high margins and succeeds, selling high-margin enterprise software will remain a closed-source business, even for vendors rooted in the Open Source community.