The net's chief architect

IS THE INTERNET running out of steam? Have we been too eager to throw in hardware to patch up network problems? And where are the thinkers in all this?

John Day argues that attempts to sort out the internet in the face of growing demands have been more about engineering and craft than science and research. We have become more vocational than theoretical.

His technological credentials suggest his claims need to be taken seriously. He shifts in his seat a little at mention of the term “internet pioneer”, but Day has been involved since the early days of the 1970s, when as a “lowly graduate” from the University of Illinois he was involved in developing communications protocols in Arpanet, the predecessor of today’s internet.

Day, now an adjunct professor at Boston University, reckons that some of the fixes that could have been implemented in those early days but weren’t, plus the lack of innovation in the years after, have been brewing to cause problems. It’s not so much about bandwidth, but deeper issues around networking, he says. “The real problems are more difficult: router table size, multihoming, mobility and security – the first three all being very closely related to failures in the architecture.”

And since the 1990s he has been developing an alternative network structure called Rina (Recursive Internetwork Architecture), which goes back to basics and makes use of a single layer that repeats over different scopes. “The resulting structure is inherently more robust,” he explains.

But reflecting on how the internet has evolved also highlights a wider issue for Day: the move away from theory, which he sees as a disturbing trend.

He says textbooks today teach you what is out there today rather than the grounding in the theories and principles behind the technology. “When I took electrical engineering I was not taught how Sony builds amplifiers, I was taught the theories of (James Clerk) Maxwell.

“Our technology has become so advanced we are mistaking craft for science,” he says. “Craftsmen are good at finding solutions to problems quickly but not so good at figuring out the next questions to ask, or asking what are these answers telling us about the nature of the problem.” He argues we tend to favour the ‘craft’ folk, while ignoring those who are good at laying the groundwork.

Speaking after his keynote address at the Science Foundation Ireland summit in Athlone earlier this month, he argues that political and business wrangling and the dominance of Moore’s Law – which maps out a trend that computer processing power doubles every roughly 18 months to two years – have not encouraged deeper innovation around internet architecture.

“We have been able to throw hardware at the problem,” says Day, in characteristic exuberance. “This often leads to an answer but not an understanding of the answer. And there’s a reticence to deviate from the answer because we don’t know where it is going to go. Getting product out the door doesn’t allow for such niceties.”

But hasn’t the internet been one of the shining technological advances of recent decades? “As long as fibre (technology) and Moore’s Law stayed ahead of internet growth, there was no need to confront mistakes. But now it is catching up on us and it can’t be fixed – every patch takes us further away from where we have to go and all we are doing is patching.”

Day is concerned over the lack of theory being encouraged in the field. “We are not taught to create scientific theory, we are taught to build things,” he says. “Two generations of researchers are conditioned and educated in this mode.”

Industry collaborations and the demands of professional advancement, such as seeking funding and publishing papers, are also skewing the talent too much in the direction of craft, he adds. “We are already seeing stagnation and it’s beginning to reach crisis level.”

Getting back to theory and asking questions will help fuel a change, argues Day. “The way you have paradigm shifts is by asking questions that the old paradigm can’t answer,” he says. “If there’s nobody pushing on that, if everyone is assuming this is the end of knowledge, you are not going to be there.”

But how do you nurture theorists in a system that tends not to select for them? It’s a tough one to answer.

“Good theorists are an odd lot, they tend to think before they build, they are more apt to listen to the problem and then pose an answer. They are a different breed of cat,” says Day, but he notes that we don’t need a huge amount of them. “A few theorists can make a lot of work for a lot of engineers – but you need a few.”

Day himself stepped back from the problem in the early 1990s to get a better perspective.

“I said ‘Let’s leave out all the political fights and constraints of hardware imposing things artificially’ . . . and as I did that, patterns started to fall out that I hadn’t seen before and I did what they told me. People say you can’t change the whole internet – but that is what they told us in the 1970s about the phone system; it’s not like we haven’t done this before.”