HoneyDanBer/NNTP/shlock(1)-style locking

Two bigs wins to this algorithm:

• Locks do not survive crashes of either the system or the application by any appreciable period of time.

• No clean up to do if the system or application crashes.

Loses:

• In the off chance that another process comes along with the same pid, we can get a false positive for lock validity.

• Not compatible with NFS or any shared filesystem (due to disjoint PID spaces)

• Waiting for lock must be implemented by polling

do_acquire ( self,  waitflag )

do_release ( self )