AARnet's first employee and the chief scientist at the Asia Pacific Network Information Centre (APNIC), Geoff Huston has warned the internet industry not to be complacent when it comes to IPv6 implementation.
APNIC is the Asia Pacific regional information registry that maintains the Whois Database, manages reverse DNS zone delegations and allocates IPv4 and IPv6 address space and Autonomous System Numbers.
Huston told Computerworld despite warnings that IPv4 addresses are running dry, there was not enough focus on developing technology around IPv6.
"There is still a lot of experimental technology going on at the moment with deeply-embedded IPv4," he said. "The industry is going to get a bit of a shock."
Internode has begun trialling IPv6 services in native mode on its national ADSL network but many Australian ISPs have no intention of running IPv6 trials in the short term despite research suggesting IPv4 addresses will run out within three years. A recent survey commissioned by the European Commission found that about 92 per cent of ISPs are either not using IPv6 or report little IPv6 traffic on their network.
Huston said Internode was doing the right thing, as the time required for training and equipment should not be underestimated.
"For years, no-one else seems to have gotten that clue," he said. "It’s a real warning about how close you have to come to the edge of the cliff."
AARNet recently celebrated 20 years of the Internet in Australia. As its Network Technical Manager, Huston was responsible for building the Internet network that connected Australian universities – Australian National University, University of Melbourne, University of Sydney, University of Queensland, University of Adelaide, University of Western Australia, University of Tasmania and Northern Territory University – between April and May 1990.
Huston is credited with providing the diplomatic approach of providing multiple protocols for AARNet. The key was the use of Cisco routers which could support TCP/IP, DECnet and X.25. As history shows, TCP/IP became the dominant protocol, but the use of multiple protocols smoothed the way and allowed the project to go ahead.
Each university had a router which acted as a gateway between the campus network and the Telstra link. Huston and Peter Elford configured the routers and travelled Australia, connecting the network. In six weeks, they had built an internet.
"I must admit that part of it was engineering and part of it was almost evangelism. It seemed what were doing was assaulting the industry monoliths of the time," he told Computerworld. "We thought it would take more than a lifetime to change anything. The word change wasn’t part of the lexicon for these companies – the services they were proposing literally came from granite. Somewhere they’d just lost their way in terms of innovation."
The real innovation for AARNet, Huston said, was the leap to a packet switched network.
"They [Telstra] were using a switching mechanism where they switched time; they had to synchronise clocks at either end of the wire and for the first half a second somebody uses it, then somebody else in the next half. It was incredibly inefficient. The only time the network was used to capacity was on Mother’s Day and Christmas Day – the rest of the time it sat idle," he said.
"We were switching packets, not time. Suddenly, everything was cheaper. It was astonishingly efficient. When I first set the network up, we leased one phone line. Admittedly, there wasn’t photos or video like today but we put the entire university traffic on it. We were 1000th of the cost of the telephone company to move that data."
The first iteration of AARNet was eventually sold to Telstra, which went on to establish the BigPond internet business. By the mid-90s AARNet had about 300 members providing internet to about 600,000 customers throughout Australia. AARNet and Telstra thrashed out a deal whereby Telstra would reduce AARNet's tariffs, introduce customer service and increase transmission speed to 6Mbps. The AARNet commercial customer database was sold for $3.8 million.
The decision to sell was hotly debated and Huston describes it today as "a mixed blessing".
"It was an extraordinarily tough decision to make but the universities found themselves running tomorrow’s carrier. It wasn’t their job and we projected, at the time, that the phone company would have to lay a submarine cable just for us over the next five years."
In fact, several submarine cables ended up being laid. Huston said although $3.8 million seemed like a lot of money at the time, looking back, it was a steal.
"Not so much the network itself as the competitive advantage of getting into the market," he said. "Telstra is such a large company that the internal fights are more important than the customers. They were mucking around for about two years beforehand. It [the sale] got Telstra into the business much faster than it otherwise would have been able to."
AARNet was able to take advantage of the distress sales of assets in the early 2000s after the dot.com boom and bust and built its own fibre network. The AARNet3 National Network was launched in 2006, giving Australian researchers high bandwidth without the previous worries about network access costs. AARNet had built more than 300km of dark fibre connecting its customers to the nearest point of the network backbone.
"To put it bluntly, in any data network you have elephants and mice," Huston explained, referring to different amounts of bandwidth used by network customers. "Most commercial outfits ended up creating a charging mechanism that works for the little users. But universities want to do big and small at once — there isn’t a tariffing system around that meets those needs. Even in 1989, there was nothing in Telstra’s book which matched what we needed — a simple data carriage environment; it wanted to sell value-added services to captured customers."
Huston said AARNet has proved that it's the applications that generate value to a network, calling the National Broadband Network (NBN) "a rare gem of the right kind".
"It is the right direction at the right point in time," he said. "Infrastructure is one of those things you constantly have to invest in but it doesn’t give you a 20 per cent return. Naturally, the private sector doesn’t want that kind of investment — nor should it."