Actually, there is reportedly a mathematical proof that you cannot
actually construct a distributed lock manager that implements POSIX
style semantics under the constraints that rpc.lockd and statd must
follow. No amount of hacking seems to have helped Sun to get them
right, and this might be the reason.

Keith Moore writes:
> Not only that, but statd and lockd have been responsible for a number
> of very crippling bugs in each of several vendor-supported operating
> systems (disasters regarding statd and lockd with IBM, DEC, Sun, and 
> HP systems all come to mind).  They are the biggest Achilles heel of
> NFS.
> 
> Which is not to say that they shouldn't be implemented, just that you
> have to be really careful how you do it.
> 
> Keith
>