It also doesn't mean Microsoft will have to go chasing comparatively small competitors in every one of those markets to compete on an even basis (as if it didn't own 80 or 90 percent of the world computer operating system market) to try to get them to all collaborate on a single, interoperable "ecosystem."
The confusion is understandable, given the requirement at Microsoft than any public discussion combine banal language and astronomically egotistical phrasing to make even the decision to visit the rest room sound like a statement of diplomatic purpose on the universality of its individual need to go pee-pee .
In its simplest terms, what Lees and other Microsofties were saying at the Partners' conference is that Microsoft is going to try to get out of its own way. It's going to try to stop competing with itself.
It's going to try to stop inventing an entirely new product to run on every new minor change in the physical layout of a computer even though the functions of the new box are almost identical and so are the functions of the old operating system and the new one.
If you make operating systems for a living, and there are five major classes of machine on which you want to run – servers, desktops, laptops, tablets and phones – it would be ridiculous to build a completely different one for each.
Wouldn't it be better if you started with a few core modules – a microkernel, a module to suck data in and spit it out, another to keep track of where it's gone, another to figure out what kind of memory or graphics or networking functions live in its immediate area and begin to adapt?
Porting those elements from one piece of silicon to another, one form factor to another, one type of device to another, is a lot easier than writing a whole OS from scratch, or buying one from someone else and trying to bolt on all the stylistic gewgaws that make it look like your OS.
If you can easily (compared to porting a whole OS) from one type of device to another, it's a lot easier to build in a hypervisor or executive-function layer of software that allows you to add whatever other functional modules you want without any of them having to talk directly to the actual hardware.
It's not quite virtualization, but it's close. Most Linux distributions are designed in similar ways, which is why it's possible for thousands of strangers to build software for it without ever meeting one another or, necessarily, knowing what any of the other development teams are doing.
Citrix' virtual-desktop suite and management apps do, too.