Summary

  • TANet’s official chronology places its first domestic peering with HiNet over T1 in January 1994; in June, SeedNet connected to TANet over 10 Mbps Ethernet.
  • The same chronology says HiNet’s connection to TANet expanded to two T1 lines around June, documenting a changing interface rather than a complete network map.
  • The records establish dates and link types, not traffic volumes, payment terms, routing policy, uptime or whether every network had equivalent reach.

The topology changed before the story acquired a map

A national network can look complete on a map and still be an island in operational terms. In 1990 Taiwan’s Ministry of Education began TANet as an academic network for education, teaching and research. Its later institutional chronology describes the first domestic peering as a January 1994 connection between HiNet and TANet in Taipei over a T1 circuit. Five months later, it records SeedNet joining TANet over 10 Mbps Ethernet. Around that point, HiNet’s TANet connection grew to two T1 circuits.

Those are unusually concrete details in a short official timeline: named networks, a sequence, a city for the first link, and two different interface technologies. They describe a transition in interconnection. The evidence does not supply a complete diagram, so the safest picture is three named networks with two documented TANet-facing connections—not a claim that the entire island had exactly three networks or only two links.

One academic backbone, two kinds of edge

TANet had begun as a publicly coordinated academic backbone. The Ministry says its purpose was to serve education and research nationwide. The University’s 30-year retrospective adds that an early backbone ran at 9.6 Kbps and that a 64 Kbps international circuit to Princeton’s JVNCnet followed in December 1991. That context matters because the 1994 links joined a network built for institutional research with two separately named services whose later accounts describe them as commercial ISPs.

The link labels matter. T1 identifies a leased digital circuit; 10 Mbps Ethernet identifies a different local interconnection technology. They are not directly comparable measures of actual traffic. A stated line rate is a capacity description, not evidence that users filled it, that the circuits were continuously available, or that their traffic was balanced in both directions. Nor does the timeline disclose who paid for each circuit, how packets were exchanged, or whether the links carried all traffic or selected routes.

The chronology’s wording calls the January link the country’s first “peering.” The 2020 university retrospective uses the same term and says Taiwan entered a cross-network era. That is meaningful institutional memory, but a modern label cannot substitute for a 1994 agreement. No exchange contract, router configuration, traffic graph or bilateral policy is cited in either account.

What the second connection adds

The June entry changes the historical picture in two ways. SeedNet’s arrival means TANet was no longer documented as connected only to HiNet. And the move from a single HiNet T1 to two T1s suggests that the HiNet-facing capacity was being enlarged at roughly the same time another domestic network joined. It is tempting to read this as proof of rapid traffic growth or deliberate redundancy. The available record supports neither explanation. The extra circuit could have served capacity, resilience, geography or another operational need; the source does not say.

The sequence still matters. Interconnection is a control surface: it determines which networks can exchange reachability through which path, and it creates practical choices about capacity, routing and dependencies. A link can make a network reachable through another network without making the two organizations equal in size, bargaining power or service scope. In Taiwan’s case, the surviving sources show the physical beginnings of a multi-network arrangement while leaving the commercial and technical terms largely out of view.

Read the record at its resolution

The Ministry timeline was updated long after the events, and the NTU retrospective was published in 2020. Both are institutional histories, not contemporaneous circuit logs. Their agreement on the broad chronology strengthens the account; it does not reveal the underlying contracts. The sources establish TANet’s academic remit, the January HiNet T1 connection, the June SeedNet Ethernet connection, and HiNet’s increase to two T1s around that time. They do not establish measured utilization, settlement, performance, routing rules, exact topology, or the reason for the upgrade.

The useful historical claim is therefore modest but consequential: by mid-1994, Taiwan’s academic network had documented domestic interconnections to two distinct ISP networks using different link technologies. That is a transition from a single-network story to a network-of-networks story. To go further, researchers would need period technical-group minutes, circuit procurement records, exchange agreements, router configurations or traffic measurements. Until such records are examined, the two links show that interconnection began; they do not tell us exactly how the resulting system behaved.

Sources

The Ministry chronology and NTU retrospective establish the sequence and stated link types. Both are retrospective institutional accounts, not 1994 circuit logs. They do not establish usage, settlement, route policy, availability, direct HiNet–SeedNet connectivity or why HiNet added a second T1.