Summary

  • The University of Minnesota’s March 1993 policy kept free use for educational and nonprofit servers that offered information freely to the Internet, while proposing negotiated fees for several commercial uses of its software.
  • The same announcement separated those terms from the documented protocol: it noted that other Gopher clients and servers already existed and could be written independently.

Historical analysis

Read the March 1993 notice slowly and its supposed blanket rule dissolves. Higher-education and nonprofit organizations serving freely accessible information were told that nothing had changed and no fee applied. A commercial company using Gopher internally was expected to pay a negotiated licence fee. If a server sold information, the team proposed a fee as a small fraction of sales, again negotiated case by case. A commercial server open to everyone sat in a grey area: a company could argue that its material delivered a broad public benefit and seek a no-fee licence.

The team had not set a universal price; it even asked users to help define a sliding scale that would not treat a small business like a large corporation.

This was a policy about using the University’s software, not a tariff for every packet or a rule that barred commercial information from the network. The post’s language was blunt about who would decide whether a public-facing commercial server deserved free use: the University. That discretion mattered. A public benefit was not an automatic exemption with a published test. It was a case to make, followed by a negotiation.

The distinction between code and protocol was explicit in the announcement itself. The team said the Internet Gopher protocol was documented, an informational RFC was nearly ready, and clients and servers had already been written by others. “You can also do it,” the post told readers. The RFC that appeared that month, RFC 1436, describes a distributed document retrieval protocol and says it is informational rather than an Internet standard. Its distribution is unlimited. It also lists available implementations and explains how new clients and servers could be built.

That does not make the University’s software irrelevant. A documented protocol can be easy to describe and still costly to implement well, port to multiple systems, support and extend. But it locates the authority correctly. Minnesota could set conditions on use of software it controlled; it could not, through that one licence policy, rewrite the protocol or require every other implementation to adopt the same commercial terms. A specification, one codebase and the people paid to maintain it were three related but distinct things.

The University’s stated reason was institutional, not technical. Budgets were being cut, the team wrote, and it was becoming harder to justify increasing development resources without a return to the University. A server that supplied useful public information also helped the campus community. A commercial service that hid its information or used it primarily to make money was harder to defend as a publicly supported activity. The policy attempted to put a price on the University’s contribution where the benefit appeared private, while leaving the public and nonprofit cases untouched.

The calendar made the comparison with the Web tempting. On 30 April 1993, CERN put the World Wide Web software in the public domain and later issued an open licence. That is a meaningful contrast in software-release policy, but it is not proof that one licensing announcement caused the Web to win or Gopher to lose. CERN’s act concerned its Web software; Minnesota’s post concerned terms for use of a specific university implementation and left room for free licences. The two decisions had different scopes and neither, by itself, records what users deployed.

Mark McCahill, the Gopher team leader, revisited the issue in a 2001 Charles Babbage Institute oral history. He recalled tight university budgets, a five- or six-person development team and a need to recover some of the cost of work that had become nearly full-time. He also remembered that a fair number of commercial licensees helped the team keep working. In the same interview, he pointed to other pressures: the arrival of Mosaic shifted control of the desktop client, and public demonstrations and conference travel competed with coding time. This is retrospective testimony, not a measured account of traffic or adoption. It does, however, make the single-cause story—“the licence killed Gopher”—hard to sustain.

The evidence establishes a negotiated licensing boundary and the team’s public explanation for it. It does not establish the fees actually paid, the number of affected servers, a fall in Gopher use after March 1993 or the effect of any one licence decision on a user’s choice. Nor does the existence of alternative implementations prove that switching was easy. The accurate historical claim is more useful than the legend: Minnesota sought a return from commercial use of its own software while preserving free use for specified public and nonprofit cases and acknowledging that the protocol could be implemented elsewhere.

That is a question of control and operating capacity, not just openness. A specification can leave future implementation choices with operators, but the choice becomes real only when someone writes, runs and maintains a working alternative. Licensing one implementation can change the economics around that work without closing the protocol. Whether that distinction protects interoperability in practice depends on adoption, portability and sustained maintenance—evidence the 1993 announcement does not supply.

Sources