WinFS wasn’t a replacement of NTFS as much as it was a supplement. Documents could be broken apart into atomic pieces, like an embedded image and that would be indexed on its own. Those pieces were kept in something more like a SQL database, more like using binary blobs in SharePoint Portal, but that database still was written to the disk on an NTFS partition as I recall. WinFS was responsible for bringing those pieces back together to represent a compete document if you were transferring it to a non-WinFS filesystem or transferring to a different system altogether. It wasn’t a new filesystem as much as it was a database with a filesystem driver.