I think this might be a little restrictive.
A NT filesystem doesn't have to operate solely in kernel mode. The kernel mode component is free to delegate work to a usermode component if it chooses. (In fact, if I'm not mistaken, Microsoft's webdav filesystem driver works this way.)
The WebDAV component of MS I know is completely in usermode. It is just a shell namespace extension.
If you meant DFS - this uses reparse points which "delegate" their work to the DFS service (i.e. usermode). That's true, but this is not a full-blown FS because of this - it's just another FS filter.
I believe boot filesystems have special requirements above and beyond other filesystems. (Like freeloader has to understand enough of it's on disk layout to load a driver off of it.)
Yap. That's surely true. Which would not be a problem in ROS (because we have the source *g*), but on NT. And I said "potentially" exactly because of this. But you cannot mount a filesystem relying on MySQL (a usermode process) early during boot process. Maybe this question should be asked in the NTFSD list - because it is about the definition of FS in NT. I just doubt that the description we currently have of this project qualifies it as "file system" in the classical sense. Nevertheless it may be a coll thing ;)
Oliver