Eh. Why wouldn't they just make a minimal OS with all the necessary hooks and then sell some first-party software that runs in userspace like everyone else instead of bundling lots of crap in the OS ;P
Every application that is bundled with Windows runs in user space; they are not actually part of the operating system.
The reason Microsoft bundles that stuff with Windows is because they are trying to undercut their competitors. Why should Dell, HP, Lenovo, and Sony pay the PowerDVD guys for DVD playback software when they could just pay Microsoft for it instead? By bundling it with the operating system, Microsoft can compete on convenience instead of soley on price/features.
The argument is predicated on an assumption, namely that "modularity" in desktop OSes will mean different commercial SKUs and Microsoft will continue splintering beyond the 4-5 different Vista models. Therefore, as people pick the SKU that is right for them the platform will disintegrate into a developer nightmare. My thoughts:
1) The splintering is already a mess. I don't know anybody who could name all of the differences between the current Vista versions, even Windows sysadmins where I work. I hope that MS sees this as the disaster it is and avoids it, although I have no data that says they do or don't.
2) Modularity in Windows 2008 represents a philosophical shift. In 2003 and before, Microsoft maintained that the advantage of a monolithic design was that it was all always there for the taking. If your .asp app needed to render a 3d model of data from Active Directory, cross referenced with an Excel spreadsheet, just suck in the COM/ActiveX objects and make it happen. Unfortunately while that is neat in concept, they took immeasurable shit from the community when core DLLs that supported stuff like that caused vanilla single task (like DHCP or print spool) servers to need rebooted monthly.
3) building on #2, even desktop OSes are becoming tremendously unwieldy. There's a valid case to be made for assuming that MS wants the ability to knife older components, partition the impact of security exploits in esoteric modules, and modernize the system. This is a lot harder than it has to be if you say it all comes in a single checkbox. Consider the Windows "Server" service exploit two weeks ago: http://www.us-cert.gov/cas/alerts/SA08-297A.html . No single PC household, virtually no multi-pc household, and perhaps a majority of commercial installations do not need this service running.
4) Painting modularity lines will allow developers to write more robust code on the platform. In the Linux world you can get a feel for the complexity or quality of a project if it's small but pulls in a lot of dependencies. A lot of cut and paste Windows coders will just google a solution and pull in whatever the first workable solution depends on. This is really easy but means that simple VBScripts and other utilities end up requiring tons of subsystems that they don't really need, taking more resources and offering more failure modes.
5) building on #4, because developers won't be able to assume people have everything installed, Microsoft will be encouraged to organize functionality in more convenient/logical ways. You frequently have to invoke ADO or WMI when using a VBScript to do simple system management tasks. While duplicating everything doesn't make any sense, it is logical to provide simple interfaces to simple tasks. This, well implemented, would mean that developers wouldn't have to learn SQL (as an unusual LDAP search syntax) in order to iterate through a user directory list.
summary: I think it's a great idea and might be their way out of the current mess of platform differentiation. They could still mess it up if they did it wrong, though.
It has always annoyed me that not all versions come with hard drive encryption. A stolen/missing laptop will benefit from encryption no matter who the user is.
