Sorry for the 20-year-old song reference, but I couldn't help myself after seeing the lines in Greg Shipley's diagram from his recent InformationWeek security article. I like what he shows but I think it can be radically more simple.
The technology world can be boiled down to two camps: those who trust their products to operate as expected and those who do not. You can guess into which camp I muster. I believe the first camp is naive and detached from the real world. (The real world is the place where intruders constantly violate assumptions, subvert logic, and make a mess of well-intentioned offerings.) The first camp spends more time talking about "enabling business" and "elevating the infosec conversation" while the second camp deals with the mess caused by the first world's ignorance of security problems.
Using this simple and intentionally provocative model I can propose two sets of lines. The first set could be labelled "compute" while the second could be labelled "transport". You could call these "host" and "network" if you like.
If you are a first camp person, the compute set is only one line -- that which computes. The transport line is also only one line -- that which transports (like a switch). This makes sense to me from a functionality (not security) standpoint. Anything of value ends up in the "OS" or the "switch". This is happening right now.
If you are a second camp person, the compute set is two lines -- that which computes, and that which verifies or at least observes the computation; call it a hypervisor or supervisor. The transport line is also two lines -- that which transports (again like a switch), and that which verifies or at least observes the transportation; call it a traffic intelligence system (to reuse terms mentioned in this blog).
This might sound suspiciously like trust but verify, i.e., trust the computer/switch but verify its operation. First campers trust, second campers trust when verified.