sylleptic - mark hopwood’s blog

I heard a terrific story at last night’s Roger Needham lecture, organised by the British Computer Society.

Mark Handley, who gave the lecture, asked his audience why the rockets on the side of the Space Shuttle are the size they are. It turns out they’re that size so they will fit through a train tunnel on their way from Utah, where they’re built. Train tunnels are the size they are because train tracks are a fixed width more or less worldwide: 4 feet 8.5 inches. Train tracks are that size because George Stephenson decided so when he built the first public railway in 1825. He settled on that size because it was the width of horse-drawn carriages, which had more or less been that size since Roman times, because that’s the width of 2 horses when put side by side.

So the rockets on the Shuttle are the size they are because of the width of a horse’s rear-end. Snopes (the popular urban myth debunking website) describes this story as “True, but for trivial and unremarkable reasons”. Mark Handley was using it to get us thinking about how things we are designing today are often constrained for the most peculiar reasons.

The Internet today is an example of that. Designed for a small number of military and academic computers to be connected together, it’s becoming the home for all data, voice, TV and mobile communications. It has real limitations around its security, how it routes traffic from one place to another without congestion and how to scale it to meet current and future needs.

We can’t just replace it by building and switching to a new, improved Internet, because there’s too much to move in one go. In the past, we’ve switched everyone’s phone number in London, we’re in the process of switching everyone in the UK from analogue to digital TV (a 10 year project in itself) but switching everything to a new, better Internet is never going to happen. So change happens incrementally, patch after patch.

Software developers know what this will eventually lead to: a system that falls apart because of the unplanned, undesigned interactions between all those patches. But change on a large scale to a better designed alternative seems impossible, so the questions that academics in this area ask themselves are very difficult to answer.

• What should the next evolution of the Internet look like?
• What’s the smartest way to migrate to that state?
• How can the change be best managed?

These questions are particularly difficult when network operators don’t really see the commercial imperative for change, when they’re so busy dealing with immediate, short and medium term problems.

For the non-engineer, there were some interesting observations about successful and unsuccessful innovation. Successful innovation generally seems to have obvious financial benefits attached to it, to be easy to implement and pick up, to be flexible enough to accommodate unexpected uses, and to be extremely well-timed. Timing seems to be the most important element for success: too early and nobody will want it, too late and something else will already meet the need, and the best solution isn’t always the one that wins.

I’ve been to 3 BCS lectures this year, and every one has given me something new to think about, but they all also deal with some common themes:

• The unplanned effects of technologies once put into large-scale use
• The complexity (generally unnoticed) of the systems we are building today
• The need to make ethical decisions about how we use and control technology

On this last point, we only have to look at how the Internet was effectively shut down in Nepal recently, and how Yahoo! have been accused of providing evidence to support persecution in China, to see how non-neutral technology has become.