If I understand you correctly Gary, you are saying if I don't set the last node in a subnet as a display flag I would cook the nodes outside of the subnet, unless I use a output SOP within the subnet then and only then would the network stops calculating once it reaches the output SOP ?
Not quite. What I am saying is that the output SOP is the explicit output node of a subnet, and not having one, houdini will default to whichever node has the blue flag on as the actual output (which could have unintended results). In short, you should always use an output SOP
I didn't know you could do representation within a subnet, I still don't understand how does adding a null between the little subnet boxes and a clip node add representation ? As well you mention without the null within the subnet, there is no way for one to see what plug is passing though your subnet, how can I see this in practice ?
By representation, I mean the ability to see the mesh of a particular subnet input from the inside, before it has been modified by a mesh inside the subnet. Try building a scene similar to the one I explained:
in a new houdini scene, add a geometry OBJ node
dive inside (you should have the default cube)
add a subnetwork SOP, and wire the output of the default file sop to the first input of the subnet
dive inside the subnet
add a clip sop and wire the first input box to the first input of the clip sop
notice how, although you can see your modified cube, you cannot see the unmodified version without exiting the subnet
now add a null node in between the input box and the clip
notice how now you can visualize the first input mesh of the subnet without having to exit the subnet