The NSFNET Backbone Services Acceptable Use Policy is dated June 1992. It is one page: a General Principle, eight specifically acceptable uses, two unacceptable uses, and a closing line noting that the policy applies to the NSFNET backbone specifically, and that networks connecting to it are expected to establish their own. The NSFNET backbone was operated under a cooperative agreement between the National Science Foundation and Merit Network, Inc., and the policy is what Merit was required to enforce. If you wanted to know in 1992 what the government-funded portion of the internet was officially for, this is where to look. The eleven clauses try to implement the General Principle.

Reading the clauses in order reveals a direction of travel. Each acceptable use extends the General Principle’s logic by one step, until by Clause 9 the policy is permitting “communication incidental to otherwise acceptable use” — which is to say, whatever is left over after everything already permitted. The two unacceptable uses arrive after nine affirmations. The document’s trajectory is not toward restriction.

Nineteen months before the AUP, RFC 1192 recommended NSF could “bar the mail and news protocols, SMTP and NNTP, from the backbone and thereby encourage private providers to offer a national mail backbone” — a stricter posture that NSF did not take. The language in the AUP — “extensive” rather than “any,” “incidental” as a catch-all, product announcements permitted “but not advertising of any kind” — was chosen by people who had seen the argument for going further.

”Open research and education”

The General Principle in full:

NSFNET Backbone services are provided to support open research and education in and among US research and instructional institutions, plus research arms of for-profit firms when engaged in open scholarly communication and research. Use for other purposes is not acceptable.

The word “open” is not describing access — it is describing the kind of research. Research whose results will be published, shared, contributed to the common record. A pharmaceutical company’s research division could use NSFNET under this principle provided its researchers were publishing, presenting at conferences, submitting to peer-reviewed journals. The moment those researchers switched to proprietary competitive work, they crossed the line. “Open” does the work of distinguishing academic science from corporate science in a single adjective.

The inclusion of “research arms of for-profit firms when engaged in open scholarly communication and research” is already an accommodation. The General Principle, in its first sentence, has made room for the corporate research lab. The theory is not simply “government and universities only.” It is “research that contributes to the open record.”

The final sentence is flat: “Use for other purposes is not acceptable.” No qualifications. The eleven clauses that follow are where the qualifications appear.

”But not advertising of any kind”

The eight specifically acceptable uses begin with cases that require little interpretation.

Clause 2 permits communication with foreign researchers provided the foreign network reciprocally allows access to US researchers — a standard for a federated infrastructure where NSFNET connected to European and Asian academic networks, each with its own terms.

Clause 4 permits use for “disciplinary-society, university-association, government-advisory, or standards activities related to the user’s research and instructional activities.” The IETF’s working groups, developing the internet’s own protocols on the NSFNET backbone, were meeting under Clause 4. The standards committees defining what the network was were using the network to do it, under the clause that specifically covered them.

Clause 8 passes traffic from other Federal Networking Council member agencies — ARPANET, MILNET, other government networks — provided those agencies’ own policies allowed it. The policy outsources its judgment for federal agency traffic to the agencies themselves.

The ambiguity is already in the list.

Clause 3 covers “communication and exchange for professional development, to maintain currency, or to debate issues in a field or subfield of knowledge.” To maintain currency means staying current in one’s field — reading, discussing, following what colleagues are doing. This is what researchers do on mailing lists and bulletin boards, and the clause correctly names it. The difficulty is that “maintaining currency” and “following interesting discussions” have no boundary the network could observe. A researcher subscribed to a mailing list that touched their field was maintaining currency in some plausible sense. The list might run wide of strict research content; the clause would still cover the subscription.

Clause 5 permits “use in applying for or administering grants or contracts for research or instruction, but not for other fundraising or public relations activities.” The “but not” structure appears here for the first time. Grant administration yes, development office solicitation no. The distinction is sensible given that NSF itself funded the backbone — it was natural to permit NSF grant applications while excluding other institutional fundraising. But the line does not correspond to any observable property of the traffic. At a research university, the same administrative systems often handled both.

Clause 6 provides the first explicit catch-all: “Any other administrative communications or activities in direct support of research and instruction.” Any other administrative communications. At a research university, the category of communications that do not at least nominally support research and instruction is a thin one.

Clause 7: “Announcements of new products or services for use in research or instruction, but not advertising of any kind.”

The “but not” construction reaches its limit here. An announcement of a product and an advertisement for that product are, from the perspective of anyone receiving the email, the same document. The distinction the clause draws is one of intent — informational versus commercial — and intent is not a property the network could observe or Merit could audit. In 1992, the products that qualified as “for use in research or instruction” were a recognizable category: scientific instruments, data analysis software, specialized workstations, academic database subscriptions. The clause was drafted for a universe where commercial software and research tools were distinct markets with distinct channels. Within three years, a commercial browser would make that distinction academic.

”Incidental to otherwise acceptable use”

Clause 9: “Communication incidental to otherwise acceptable use, except for illegal or specifically unacceptable use.”

“Incidental” is a precise word here. It does not mean accidental or merely tangential — it means arising in the course of something else: communication that happens because you are already doing something that is permitted. An email to a collaborator about when to call is incidental to the research it supports. A departmental mailing list that ranges off-topic is incidental to the professional relationships it serves. Personal correspondence between colleagues who are also friends is incidental to the academic network those friendships are embedded in.

This is what most of the personal communication on academic networks was. The policy needed Clause 9 because a network for researchers cannot run without the interpersonal traffic that researchers generate in the course of working. Remove “incidental” and the policy prohibits most of what makes research institutions function. Leave it in and you have a catch-all whose outer boundary is whatever “incidental” means in a given case.

No enforcement mechanism existed for Clause 9. Merit was operating infrastructure, not auditing the purposes of individual messages. Nobody was reviewing whether any given communication was incidental to otherwise acceptable use or simply personal traffic that found the academic network convenient. Clause 9 is not a limit — it is an acknowledgment. The network was already being used this way. The policy was not going to pretend otherwise.

The three catch-alls — Clause 3 (professional development, maintaining currency), Clause 6 (any other administrative communications in direct support of research and instruction), and Clause 9 (anything incidental to acceptable use) — overlap in a way that covers most of what people on the network were actually doing. Between them, the acceptable use clauses describe a network with a wide center.

”Extensive use for private or personal business”

The two unacceptable uses, quoted in full:

Use for for-profit activities, unless covered by the General Principle or as a specifically acceptable use.

Extensive use for private or personal business.

Clause 10 is a cleanup clause. It restates what the preceding acceptable uses have already established: for-profit activities are prohibited unless they meet the acceptable use criteria. True by definition, placed here for formal completeness.

Clause 11 is different. The key word is “extensive.”

“Any use for private or personal business” would be an absolute prohibition — a clear rule, even if difficult to enforce. The standard would be unambiguous: no personal use, period. “Extensive use for private or personal business” creates a spectrum instead. Some personal use is permitted. A large amount is not. The line between them is not defined.

By June 1992, personal use of the NSFNET-connected internet was ordinary among the population with access to it — graduate students, faculty, and research staff at universities. It had been ordinary for years. The academic internet was where those people communicated. “Communication” included personal communication, had always included personal communication; the network connected people, and people are not segmentable into professional and personal functions at the packet level. An absolute prohibition on personal use would have required enforcement against the network’s established users, which NSF was not going to undertake, and which Merit had no mechanism to carry out.

“Extensive” is the policy choosing a word that accommodates the reality while preserving the official posture: this is a network for research and education; personal use is tolerated as long as it does not dominate your use. What counts as “extensive”? The document does not say. No threshold was defined. No audit mechanism existed to distinguish extensive from incidental personal use. The word did what it needed to do: it placed something in the unacceptable column while ensuring the prohibition was practically unenforceable.

In 1995, NSF decommissioned the NSFNET backbone and transferred the infrastructure to commercial operators. The AUP ceased to apply, not because it was formally repealed but because the network it governed ceased to exist in the form the document described. The available sources do not establish whether any document formally terminated the policy.

“Extensive,” “incidental,” “but not advertising of any kind”: the eleven clauses were the last written statement of what the government-funded internet was officially for.