That circuit was the best in terms of least components, but it does not work well if your load has capacitors across the supply rails. The one I did back on the first page of posts has one extra resistor, but it is the one that works best with a capacitive load. The circuit is from LTSpice, so there is extra junk in the diagram for the simulation.
To me the method in general is not a robust solution as it depends a so much on the way the load behaves. For example, a transient from the load circuit can turn the supply on, even if the switch is not touched. If your load was a linear regulator IC with a reverse diode across it, and you attach an external load to the regulator output which happens to have a charged supply cap, then this switch circuit will turn itself on
Richard.