With open-source software businesses, you have two options. Actually, three, but the third belongs to Red Hat, and it applies to roughly no one else.
The first option is to sell support for open-source software. This option is generally advocated by those who have never grown a business beyond $10 million. It's a terrible model unless your only aspiration in life is to run a services company.
Hence, the support model might be good for Accenture or systems integrator, if they want to take on the burden of support, but it's a poor model for Red Hat, MindTouch, Microsoft, or other software company.
The second option is to contribute heavily to open source but not build your revenue model around monetizing that software directly. This is what the New York Times points to in its Sunday expose of allegedly fizzling open-source business models.
Open source can drive adoption like little else. It's not, however, necessarily a great driver of revenue. For that, you need to be selling something beyond the source code, and that "something" is often going to be proprietary, be it hardware, software, or a service.
Google is the master of this model. It gets roundly criticized for its half-open, half-closed open-source efforts, but the reality is that Google's products--Chrome OS, Android, etc.--are open enough to facilitate adoption without giving away the keys to the car, which drives wherever Google wants it to go.
That's the way successful companies are run: they take ownership of what they ship. They are influenced by but not controlled by the mystical whims of The Community.
Even Red Hat, which piggybacks on a lot of Linux kernel development, increasingly includes more home-grown software in its distribution and takes great pains to certify its Red Hat Enterprise Linux will work in the most demanding environments before putting its brand on the label.
Some, including me, have wrongly concluded that Red Hat's business model would apply to other product markets beyond the operating system. It doesn't. It only applies where the moving parts in the product are complex, multitudinous, and frequently changing.
For everything else, there's Option 1 (if you want a business that doesn't scale well or possibly at all) or Option 2 (which is really no different from the old proprietary model except that it effectively uses open-source complements to lower engineering costs and possibly sales/marketing costs).
Even Option 2 won't work if you under-invest in marketing and development, as Symbian is learning to its hurt. It turns out that there is no free lunch, even in the land of free software. It always takes work. And money. Lots of both, actually.