Canberra became the capital of Australia for the same reason Washington, D.C. became the capital of the U.S. In Australia’s case, Canberra was chosen because it wasn’t Melbourne and it wasn’t Sydney and because it was somewhere between the two.
Canberra is also the capital of the Australian Academic Research Network (AARNet), a network managed by Geoff Huston. The network came into being after ACSnet began to feel the strain of, as Geoff puts it, “networking on two bob.” ACSnet was strictly a volunteer effort between the computer science departments, with no management and no funding.
DEC’s educational group had been providing international access, but after somebody took over the pipe for two days for a massive file transfer, they pulled the plug. Robert Elz, who had been handling most of the international liaison work, arranged an X.25 connection to UUnet in Virginia.
X.25 costs money, so Elz wrote software to provide accounting and chargeback to the users. Then, the first leased line, operating at 2,400 bps, was installed between Melbourne and Sydney. Networking was starting to cost money and it was evident that a more coordinated effort was needed.
About this same time, the central computer departments decided that networking shouldn’t be left to computer science. The computer departments put together a proposal for a national research network based on X.25 and the U.K. Coloured Book standards.
This group went to the federal government, but the feds had no interest in funding a national network. Next, they approached the committee of Vice-Chancellors, a group of the heads of all 38 universities formed to lobby for the interests of higher education against other groups competing for a slice of the federal pie.
The Vice-Chancellors were interested, but weren’t about to just hand the computer centers carte blanche. They initiated their own feasibility study. With the feasibility study, they decided the idea was worth pursuing and, in January 1989, formed two groups to develop the idea.
The first, the technical group, met for the first and last time that same month. “It was bloody obvious what needed to be done,” Geoff said. The network would simply be multiprotocol routers and leased lines. No code to write, no research issues to investigate. You simply took out the price books and tariff charts and figure out what you could afford.
The policy and steering group only met three times. The Vice-Chancellor who chaired that group made it clear that they were there to build a network, “so let’s get on with it.” Geoff Huston is credited by many in Australia for helping to make sure that the process moved smoothly in both the policy and technical groups.
By November 1989, everything was done. Geoff placed orders for leased lines and Cisco routers, telling both vendors he wanted a delivery date of April. He told Cisco how he wanted the routers pre-configured. With that much lead time, there was no excuse for delivery problems and none occurred.
In the last week of April, all the boxes were in their locations and the 48 kbps lines were up and running. For the next three weeks, Geoff and his one other staff member toured the country turning on the boxes and connecting them to the campus Ethernet and the leased lines. In a total of three weeks, the entire network was up and running.
The network continued to be run with only two people, even though it had expanded to 2 Mbps lines on key routes and had a huge growth in traffic. By January 1991, over 40 Gbytes per week were being transferred on the network. One year later, 77 Gbytes per week went through, and peaks of 120 Gbytes per week were not uncommon.
The lack of religious battles over the network and the clockwork installation are widely credited to Geoff Huston’s deft leadership. He defused any attempts to make the network overly complicated and insisted on a multiprotocol backbone so that any traffic the customer wanted to transfer could be transferred. He showed that installing a national network can truly be no fuss, no bother.
While hooking up the network internally was straightforward, there still remained issues in the Internet as a whole that were having a real effect on the operation of AARNet. Geoff gave me a dramatic example of one of these issues.
He started a traceroute to Barry Shein of Software Tool and Die, a well-known member of the Internet community. The traceroute went out from Canberra to the AARNet routers, reached NASA Ames and then failed. Barry is connected to the Internet via one of the Commercial Internet Exchange (CIX) members and AARNet connects to the Fix West. Since the NSFNET had refused to allow its routers to announce the presence of networks that had not signed the Appropriate Use Policy (AUP), the result was a fragmented Internet.
Software Tool and Die was a classic example of commercialization of the network. Barry Shein started the operation as a commercial IP service provider, giving over 1,000 people access to the Internet for an hourly fee, exactly the kind of thing the AUP warned against. Software Tool and Die could have waffled the issue, signing the policy and issuing severe warnings to its users, but that seemed kind of silly.
Since Software Tool and Die was an AlterNet hub, Barry Shein estimated that they had access to over 80 percent of the Internet. AlterNet was on the CIX, and most regionals had set up back door paths to one or the other of the commercial networks. Several commercial links to Europe existed, making international access to the EBONE and its regionals straightforward. In the U.S., only a few regionals in the Midwest had a single connection to NSFNET and were unreachable by Software Tool and Die customers. Most of the Pacific Rim, however, squeezed through the single Fix West link, and was cut off.
Geoff Huston was quite adamant that the Internet was facing many serious issues, fragmentation being one of the more visible. The address space was quickly filling up, there were so many networks that routing protocols were thrashing, and a host of other problems were surfacing.
These problems were not inevitable in Geoff’s eyes. The Internet began as a single global community with ad hoc regulation of the network for the common good. Over time, under the pressures of massive growth and commercialization, the common view splintered into many camps. Global connectivity is a key issue for the Australian community. At the end of a long, skinny pipe, much of their traffic is communication with other parts of the world.
There is no group that worries about global connectivity, or even provides much in the way of a global regulatory framework and places like Australia, on the frontier of the Internet, feel the impact of this lawlessness first. The IAB had long been charged with looking out for the Internet, but Geoff pointed out that the heavy American bias of the IAB doomed it, particularly in Europe. Many Europeans continue to refer to the TCP/IP protocol suite as “DoD IP” to emphasize the American military origins of the Internet.
While the IAB had failed to provide a global framework for connectivity, the engineering arm, the IETF, had not always provided solutions to the problems that needed solving. Geoff emphasized that the IETF was a “tremendous asset” but also cited Marshall T. Rose’s comment that “adult supervision” was needed.
The problem he saw was the lack of direction given to the IETF. The IAB and the IETF Steering Group (IESG) failed at times to provide a solid agenda and timetables of engineering problems that needed solutions. Problems such as the address space exhaustion were studied in a fairly ad hoc fashion, even as the problems became more and more pressing.
The IETF also suffered from an influx of goers and an increasing tendency by many to treat the occasion as a social forum rather than a working environment.
So what forum should guide the evolution of the Internet? Geoff placed some hope in bodies like the Internet Society, but also wanted to see an operational solution to the problem of global connectivity.
To participate in the global telephone network, you have to agree to certain sets of regulations in order to plug in. Geoff wanted to see some baseline regulations defining how one cooperates in the global network and, most importantly, a place to plug in to a global, neutral backbone.
The NSFNET is an American backbone that also doubles as a global backbone, provided you play by NSF rules. Geoff wanted to see some higher tier, say some FDDI-like point of interconnection. This ring would have no policies prohibiting certain classes of traffic. If you pay your share of the ring infrastructure, and abide by certain technical rules, you can join.
Once on the ring, it is up to each individual member to decide whether or not to accept traffic from other ring members. This bilateral decision would allow the NSFNET to decline to accept commercial traffic without at the same time fragmenting the connectivity of the entire Pacific region.
That night, over dinner and drinks at Geoff’s house, I brought up the question of international bodies again. He had high hopes for the Internet Society and I was curious whether he shared the opinion of Bob Kummerfeld in Sydney that the OSI process was a negative influence.
Geoff felt that OSI was a good thing. It had kept all the goers busy, preventing them from turning their attention to forums where real work was being done. Had OSI not existed, all these people would have looked around for something to do and found things like IETF meetings.
Under the Huston theory of standards development (and like any good theory, it helped predict and explain many phenomena of the standards world), one of the best things that could be done for the Internet community would be to initiate a movement for OSI++, keeping the standards potatoes fully occupied for another ten years.