All my memory is not firing these days (somtimes working on 71 pins), but methinks I recall from my active duty days a requirement that came out around 89/90 (?) that stated something like "all future procurements of DOD computer hardware and software (in 1980's verbiage, I'm sure) must be OSI compliant". I remember shaking my head, wondering where and how I was going to find this equipment, but it never became a problem because there was (may have also been) a clause that included "TCP/IP" as an acceptable option ... I just don't recall all the details.
I think this may have been driven by the reasoning of some at the time the we would soon be seeing TCP/IP in our rear view mirror was we embraced "International Standards" (CMIP anyone?).
Further background ... the "new" DOD messaging progam (Defense Message System, or DMS) RFP included a requirement for the selected vendor to offer software that supported both "Internet" protocols and "OSI" protocols. In all my contact with military commands and civilian agencies working in DMS implentation I have never come across anyone who was running OSI protocols in house, and would therefore need (or choose) the OSI version of the software. Also, DMS requirements state that if you do use the OSI version of the software, you must translate CLNP to IP via your _own_ "gateway" into the uNclassified but sensitive Internet Protocol Routed NETwork (NIPRNET), or Secret Internet Protocol Routed NETwork (SIPRNET), because those military networks (and their associated "NAPs", if you will) only - by definition - support IP today.
Note "military" references above. I wonder, then, if the definition of a "NAP" associated with this requirement refers to an ARPANET NAP, (look up "legacy" for a picutre) and really has (or should not have) any meaning today. Perhaps it is time to really redefine a NAP.
On my own