Summary

  • Merit’s 9 April 1993 update scheduled an NSFNET policy-routing database addition for RO-EARN, but the record is not evidence that a transatlantic packet completed a round trip.
  • ICI/RO-TLD dates Romania’s full connectivity to America and the wider world to 16 April; its retrospective does not reproduce the original status-change message.

The most revealing Romanian Internet milestone of April 1993 may be a small row in a routing-policy archive. The entry is easy to overread. It lists an address block, a Romanian network name and preferred routes. It does not show a packet moving along those routes, a user reaching a host abroad, or traffic succeeding in both directions.

The archived update is dated 9 April. Steven Widmayer of Merit wrote that the additions would be installed on the NSFNET backbone by 08:00 Eastern time. The row reads 192.162.16 RO-EARN C:RO D:04/09/93 1:513 2:1800; the accompanying network details identify RO-EARN in Bucharest and route priorities through EASInet/CERN and ICM-Atlantic. This is direct evidence that a policy record was being changed and scheduled for installation. It is not a measurement report.

That distinction matters because the institutional chronology puts a consequential change one week later. ICI and the Romanian top-level domain registry say the institute received an email from Dr Steven Goldstein on 16 April after the country codes RO and BG were reclassified to status W, described in the retrospective as “most favoured nations.” ICI says NSFNET then permitted traffic routing for Romania and Bulgaria, and identifies that date as full connectivity to America and, through America, the rest of the world. The public account does not include Goldstein’s original email or the underlying classification record.

The two sources need not contradict each other. A policy table can name a network and preferred path before a later restriction is lifted or before a route is usable under the governing policy. Conversely, the 9 April row alone cannot tell us whether the listed paths were already carrying any traffic. The strongest conclusion is narrow: the archive records a planned policy-database addition; ICI’s later account identifies 16 April as a policy and connectivity milestone. Neither source is a packet capture.

Other dates add context but do not collapse into a single “first connection.” Romania’s .ro domain was registered internationally on 26 February 1993. That administrative milestone did not establish global host reachability. A May 1993 RIPE report says a direct 14.4 kbit/s IP line from Bucharest to Vienna had operated since March and carried both Internet and EARN traffic; it also describes EARN as the most active Romanian network and notes that a dozen other institutions dialed into EARN nodes.

ICI’s later timeline describes a 9.6 kbit/s Vienna line for EARN in 1992, while another retrospective dates a permanent ICI–Vienna line to December 1992. These may describe successive services, but the sources reviewed do not fully reconcile them.

The evidence therefore supports an academic and research-network story, not a claim that Romanian households suddenly received public Internet service on a particular April day. It also illustrates why a database object, an international circuit and an operating route answer different historical questions. A domain registration identifies a name; a leased line describes a physical or service connection; a routing-policy row expresses how an operator intended traffic to be selected; successful reach requires operational evidence beyond all three.

Sofia Ren’s earlier account of Poland’s 1991 connection follows a different trail: a Warsaw TCP/IP link carried over a borrowed BITNET circuit and a later move from European to global access. Romania’s story here is narrower. It centers on a route-policy entry, a seven-day gap, and the careful distinction between configuration and demonstrated reach. That gap is not a flaw to smooth away; it is the limit the surviving evidence gives us.

Sources