It sounds like he's asserting that userspace filesystems just aren't capable of the same performance as kernel-mode filesystems. What kinds of constraints make this true? Are they inherently so, or is it just that Linux is designed one way now and we'll be damned if we have to change it?
ntfs-3g uses fuse, and even though I wouldn't run my root fs on ntfs under linux, I've found that it meets my needs quite well.