The back-off on use of .Net in Vista/Longhorn has been well known and publicly acknowledged for about a year and a half, for anyone paying the least bit of attention to Longhorn/Vista development.

Reasons given have been the following:

- Work on XP SP2 left fewer programmers available for work on Longhorn. Code base was far behind coding and QA work done for SP2. They started over with the latest branch and reintegrated their Longhorn work as it became stable enough.

- Too many new subsystems interdependent on each other. The managed shell relied on .Net 2.0, WPF(Avalon) and WinFS. All of those were managed code in heavy development. WinFS won't even be done in time for Vista's ship date. Do you keep fixing your code to keep up with API changes (WPF was changing quite a bit from CTP to CTP until recently) or build those libraries as standalone install-ables (with the added bonus of being able to ship them on the previous OS to build install base) that can be either bundled with Vista or shipped when ready. And at the very least WinFX will be on the hard disk as an MSI installable from the "Windows Components" control panel. More recent word is that it will now be installed by default as well. They say they were worried about the increased attack surface of the API in a default install.

- Already late. Need to ship something in reasonable time frame. Do you write a new shell from the ground up in a new language or build something with most of the same features on the existing code base? I think the answer is somewhat obvious.