With GNU/Linux you're stuck with "classical" filesystems. GNU Hurd allows to have all sort of fancy things, for example to file to be both file and directory (which is totally impossible in Linux - and Linus strongly opposes any attempts to change this unfortunate state of things).
And it matters a lot to typical everyday GNU/Linux users. They frequently struggle (I, personally, do) with GVFS/KIO kludges (and other non-kernel "filesystems") - which GNU/Linux systems must have because even FUSE filesystems can't generally adequately replace them.
Microkernels in general make for a great case-study on why arbitrary flexibility may not provide user value. Most of the OS doesn't have a need for "alien," radically innovative subsystem development - improvements to things like scheduling are mostly incremental in nature. Filesystems are one of the huge exceptions to this rule.
A good example, because GNU Hurd has different view on filesystems implementation. Debian GNU/Hurd project has fairly nice explainations here: http://www.debian.org/ports/hurd/hurd-doc-translator
With GNU/Linux you're stuck with "classical" filesystems. GNU Hurd allows to have all sort of fancy things, for example to file to be both file and directory (which is totally impossible in Linux - and Linus strongly opposes any attempts to change this unfortunate state of things).
And it matters a lot to typical everyday GNU/Linux users. They frequently struggle (I, personally, do) with GVFS/KIO kludges (and other non-kernel "filesystems") - which GNU/Linux systems must have because even FUSE filesystems can't generally adequately replace them.
Edit: However, this not as flexible as Plan B's boxes (http://plan9.escet.urjc.es/ls/export/box.html).