> Also comparing the two subsystems is entirely wrong. ReadDirectoryChangesW is still a polling mechanism, inotify and dnotify are kernel level events.
ReadDirectoryChangesW sleeps while there aren't any changes and only returns once there are some changes. Efficiency-wise there isn't any real difference between kernel-level events and sleeping on a different thread, and I don't see one as clearly better than the other. After all, waking a thread up is also a kernel-level event!
> Not only that but you still need a password for single user mode, and it would take a bit longer for him if he had a boot disk.
Who cares? The difference in the level of security you get is a small epsilon. You either don't care about physical attacks, in which case you don't care whether it takes 5 seconds or 5 minutes to break, or you do care about physical attacks, in which case you encrypt your data.
Of course, your entire point is irrelevant -- calling Windows 9x "on the top of it's [sic] game" in terms of security is lying. Windows 9x was insecure in much more serious ways than a stupid login prompt. Windows NT-based systems, especially Vista onwards, are much more effectively designed.
> This isn't about Microsoft Research, this is about Microsoft products. This is a completely irrelevant point.
You used the words "success" and "innovation". The fact that MSR exists and is as good as it is, is an important part of Microsoft's success as a whole, and there's no lack of innovation on display there (witness Street Slide for a recent example). To consider "success" and "innovation" only in terms of released products is myopic.