Summary

  • “Australia joined the Internet” compresses several milestones: earlier research email networks, the Melbourne–Hawaii TCP/IP exchange in June 1989, an international circuit, and AARNet’s later national rollout.
  • The University of Melbourne reports 2,400 bps for the first connection; AARNet’s anniversary history reports a 56 Kbps satellite circuit. The sources do not say whether those figures describe the same measurement layer, so the difference remains unresolved.
  • The connection mattered because it made a live TCP/IP path available at Melbourne. It did not begin Australian computer networking, and it did not put every institution or household online at once.

Two speeds for one beginning

A 2,400-bit-per-second connection and a 56-kilobit-per-second circuit sound like two incompatible descriptions of one event. They may refer to different points in the path, but the sources do not establish that. A careful history has to keep both statements attached to the source that made them.

The University of Melbourne’s computing history says the first TCP/IP Internet connection to Australia opened to the university on 22 June 1989 in Hawaii, which was 23 June in Australia. It identifies Robert Elz as the person who conducted the connection and says the first inbound message carried the subject “Link Up”. The same page reports a speed of 2,400 bps.

AARNet’s commissioned twentieth-anniversary book tells a different part of the story. It calls the Melbourne–Hawaii connection a 56 Kbps satellite circuit, completed by Elz and Torben Nielsen. A 1999 Age anniversary report also describes a 56 Kbps stream, carved from a 64 Kbps link, and recalls technical work to match the endpoints. Geoff Huston’s 1996 presentation, written while the network’s early history was still recent, also lists a 56 Kbps Melbourne–Hawaii link.

Those accounts agree on the broad event: a working international Internet path reached Melbourne in June 1989. They do not agree on the number attached to its speed. A satellite circuit’s rated capacity, a connection as experienced at a host, and an application’s observed rate are different possible measurements. That distinction could explain the gap, but no source consulted here says it does. The University page does not name a measurement point; the retrospective accounts do not explain the 2,400 bps figure. There is no basis for choosing one number as the definitive rate.

Even the dates need a scope label. The University page explains that 22 June in Hawaii was 23 June in Australia. The Age account places the working connection overnight on 23–24 June Australian time. That is compatible with a connection established across time zones; it does not prove a single exact timestamp for every stage of the circuit and host session.

Australia was already communicating

The year 1989 marks an important change in Australia’s connection to the Internet. It does not mark the beginning of computer networking in Australia.

The National Film and Sound Archive’s account of “Australia Dials Up” traces university and research networking to the 1970s. Computer-science departments used telephone lines to exchange files and messages; ACSnet later connected Australian Unix hosts. By the early 1980s, the archive says, a permanent Australian email connection to ARPANET had been established. These systems carried useful research communication, even though their paths and timing differed from a continuously available TCP/IP connection.

A 1993 ANU account describes how national networks such as BITNET, JANET, EARN and ACSnet connected researchers in different countries. It says that by the end of the 1980s AARNet provided professionally managed access to such services in Australian universities and research institutions, with overseas interconnection through the Internet. The wording preserves a transition: users had networked services before the national Internet link, and AARNet connected those services to a more continuous international path.

The Australian Unix Users Group’s 1992 ACSnet survey offers a closer view of the overlap. It calls ACSnet a network linking many Australian Unix hosts and lists connections to AARNet, the Internet, Usenet, CSnet and overseas networks. The survey also says ACSnet remained available to commercial sites after AARNet formed. That is not evidence that all of these networks were equivalent, nor does “only such network” in the survey mean that Australia had no other computer networks. It does show that the older system persisted alongside the new one.

The word “Internet” therefore needs care. Before the Melbourne–Hawaii exchange, Australian researchers could send messages or files beyond their local computer using other protocols and scheduled paths. The 1989 event was different: it established a live TCP/IP connection to the international Internet. Calling this the first Internet connection is defensible when the claim names that technical scope. Calling it the first Australian network or the first Australian electronic communication is not.

A host, a message, and a circuit

The first message’s short subject, “Link Up”, is an unusually good label for the event because it is also unusually small evidence. The University of Melbourne’s history identifies the subject; a 1999 report says the phrase was probably “Link up …”. The wording is remembered through later institutional and journalistic accounts. It confirms the symbolic moment without supplying a complete transcript of every packet, route change or circuit test.

Melbourne was a practical endpoint. The University’s account credits the work of its computing department and Robert Elz, who had helped build Australian network and naming services. AARNet’s own history says the first Australian Internet connection was built at Melbourne with a direct connection from the University of Hawaii. These descriptions place the endpoint where technical staff, an existing host and international research links met. They do not make the connection the work of one person alone.

A host-to-host TCP/IP exchange was also not the same thing as a national backbone. AARNet’s current history says the Internet feed was extended to Australian capital cities and the Australian National University in Canberra after the first connection. Roger Clarke’s history dates the full rollout to Australian universities and CSIRO to April and May 1990, carried out over four weeks. AARNet’s twentieth-anniversary book also separates the 1989 trans-Pacific link from the network implementation that followed.

The difference changes how “connected” should be read. Melbourne had an international path first. More institutions joined later, through a backbone that had to be built, staffed, and connected to local networks. An account that uses the June 1989 date to imply immediate national or household access gives one event more reach than the records support.

The permanent connection also changed the rhythm of communication. The University of Melbourne page quotes academic Lee Naish remembering that before the permanent link, one would rarely be online at the same time as the person one wanted to contact. A live path reduced that timing constraint for people who could use it. It did not make bandwidth large, coverage universal, or every service interactive. Email and Usenet remained central ways to share information, and the first connection was small by later standards.

The first is a boundary, not a birthday

Histories often need a date, and June 1989 is a useful marker. Yet a date without a defined event can turn several milestones into one origin story. The Australian record distinguishes at least four questions:

  • When did Australian research communities first exchange messages? The histories point back to earlier telephone, UUCP, and ACSnet routes.
  • When did a TCP/IP connection reach an Australian university from the international Internet? The University of Melbourne identifies the June 1989 Melbourne–Hawaii session.
  • What was the capacity of the first international circuit? Retrospective accounts report 56 Kbps, while the University page reports 2,400 bps for the first connection; the public sources do not reconcile them.
  • When did a national research network connect universities and CSIRO sites? AARNet’s rollout followed in 1990.

The answers describe related but non-identical states. A message route can work before an always-on circuit. A host can exchange TCP/IP packets while most institutions remain outside the backbone. A national network can operate before the public has broad retail access. The historical record becomes more useful when it keeps those states separate.

This distinction is also a matter of evidence. A commemorative page can preserve a milestone, while a surviving email or technical log may establish what ran and when. The Melbourne history points to extracts from email recounting the 1989 permanent connection, and the 1999 newspaper account attributes details to people involved. The sources available here still leave the rate discrepancy open. A better archive could resolve it; a confident retelling cannot.

The accurate short version is not that Australians had no network before 1989, nor that the whole country came online in one night. It is that Melbourne gained a live TCP/IP connection to Hawaii in June 1989, a step built on years of Australian research networking and followed by AARNet’s national expansion. “First” becomes precise when the layer is named.

Sources and evidence limits

The core contemporary and institutional records are the University of Melbourne’s early Internet history, AARNet’s institutional history, the 1992 Australian Unix Users Group ACSnet survey, and ANU’s 1993 account, “Revolution in Research”. Retrospective evidence includes AARNet’s twentieth-anniversary book, Geoff Huston’s 1996 APNIC presentation, Jenny Sinclair’s 1999 Age anniversary report, the National Film and Sound Archive’s “Australia Dials Up”, and Roger Clarke’s history of AARNet’s rollout.

These records support the sequence and the published claims about rates. They do not expose a complete 1989 circuit log or explain whether 2,400 bps and 56 Kbps measured the same thing. The article does not infer a traffic volume, adoption rate, or household availability from the first link.