state of the nfsv4 server

[J. Bruce Fields]

This is my current list of things I'm expecting would get fixed before
removing the "experimental" marking from the nfsv4 server code.  This
isn't a complete bug list--it's just the things that I think of as being
especially big, or that may involve backward-incompatible interface
changes that could affect users.

	- export paths consistent with NFSv4: right now we require you
	  to construct export paths in an odd way (with fsid=0, bind
	  mounts, etc.), and paths aren't consistent with the paths seen
	  by NFSv2/v3 clients; more details at
	  http://wiki.linux-nfs.org/index.php/Pseudofilesystem_improvements
	  but we have no code right now.  We could definitely use help
	  here....

	- reboot recovery: our reboot recovery code needs to be
	  replaced.  Some of it will be replaced by a userspace daemon,
	  so the transition will require people to update nfs-utils in
	  tandem (ouch).  We have a prototype; with luck it may be ready
	  for 2.6.21.

	- security on exports: we'd like to allow export options like
	  sec=krb5:krb5i:krb5p instead of using the special gss/krb5
	  client names in the exports file.  This will also allow
	  specification of different security requirements for different
	  ip address ranges (e.g. if you trust a local network), which
	  is something a few people have asked for.  And we'll implement
	  security negotiation at the same time.  We have prototype
	  code, which should be ready for 2.6.21.

No doubt I'm missing something; let me know.

--b.