jhuns@vt.edu:
the sole problem, from my analysis of the wide variety of aup available from around the world, is that each governs part, but not all of the network after a certain point of plurality. because of this, even though we can speak of an nsfnet aup, it may have been possible even that time to use nsfnet without being necessarily bound by that policy because of interconnections between networks
I think maybe this collapses two different things: (1) When the NSFNET was the One True Backbone -- and governed "all of the network" -- networks could not interconnect to it and legitimately pass commercial traffic. Some commercial traffic was passed of course but no network provider wished to connect to NSFNET as a business model. This preserved a non-commercial character but stifled growth, and is why the NSFNET AUP was important in functional (organizational/institutional) terms although naturally yuo could find many individual traffic streams which didn't respect said AUP on a usage basis. Some network providers saw a commercial opportunity in Internet though and created CIX. CIX danced around NSFNET so that NSFNET no longer governed "all of the network", 1991. Then NSFNET declared that it was no longer responsible for the whole backbone, only one of them, 1992, and got out of the way entirely, 1994. So, between 1991 and 1994 the Internet progressively became a set of interconnected networks abiding by common standards instead of a hierarchical system. By serving as the initial technical interface for non-NSFNET-AUP inteconnection, CIX was the turning point in the Internet's passage from a narrowly-used research network to a widely-used commercial network. (2) Once the NSFNET replaced itself with 4 Network Access Points (NAPs), alongside CIX, and hence consecrated the Internet's many-interconnected-networks architecture, indeed the only relevant traditional AUPs became the Acceptable Use Policies that the majority of networks impose upon their users (the "wide variety of aup available from around the world" which you have analysed). But note also a third element which maybe brings us back to where this discussion started: (3) There continues as you know to be another, more diffuse AUP-like device that governs the Internet -- Internet standards. Collectively, the Internet's corpus of RFCs (Requests for Comments) and, specifically, the BCPs (Best Current Practice documents) which award some of them special status, constitute today's version of the NSFNET AUP (rfc-editor.org). Those who do not obey the RFCs open themselves to losing interoperability. Like any policy document however enforcement of the RFCs/BCPs -- copyright still held, I think, by ISOC (isoc.org) -- is diffuse and somewhat inconsistent. Especially so on the Internet where said enforcement is collaborative not centralised. Hence non-compliance means, again, likelihood of less interoperability (outcomes of provider-by-provider decisions), not black-or-white on-net or off-net status. Social sanctions not legal sanctions comprising governance whose institutional and informal materialisations are a quite valid research subject, as socio-legal scholars might chime in. my C$0.02, Bram