Bulletin 14 June 2019. Why do we assume the Internet will ‘just work’?
The connectivity disconnect.
I was recording a webinar yesterday. For the uninitiated, this is the closest I will ever get to being on telly: an hour of sitting around a table chatting in front of cameras, to be broadcast to the world. Well, narrowcast: the audience tends to number hundreds rather than thousands, but heck, it feels like what telly might be like. They sometimes even have makeup.
It’s amazing what can be done this days, with a few HD cameras, skilled operators and a network connection. What’s that you say, a network connection? Yes, you know, that bit that absolutely has to be there, otherwise nothing can work. I know, obvious, right? Less obvious is why we tend to take the network so utterly for granted.
The Internet is a clever beast. It’s built on the back of protocols that assume the worst: data packets can be re-sent if not received, or routed in multiple directions in case of failure. Like Kevin Costner’s futuristic (yet ill-fated) postman, the message has to get through. Yet the way we have configured things, we generally connect across a single wire — making for a single point of failure.
To whit, the webinar had to be postponed (for only minutes, in the end) as a ‘network problem’ was resolved: no connection, no webinar. I’m not saying the world needs redundant (as an aside, which idiot decided that the same word would mean both extra and superfluous?) networks; rather, I can pretty much guarantee that nobody gave the network a moment’s thought up until that point.
After all, it should just work, shouldn’t it? That certainly seems to be the thinking of the many metro-Californian types I seem to speak to, who present tools such as Google Docs as a straightforward alternative to desktop office applications. I don’t hate Google Docs. However… (and yes, this sounds like a moment where I now reveal my prejudices)… however, I would like it a lot more, if it worked in environments where network connections were sporadic.
Such assumptions pervade the thinking of a large number of people for whom infrastructure is ‘just’ a platform (a.k.a. anyone who works in software, sorry guys). The network operates quite differently to other infrastructure elements: in that golden triangle of processing, storage and communications, the relationship between the first two and outcomes tends to be linear — the more you have, the better the result.
For the network however, the relationship is both linear and binary. Scale it up and down by all means, but also act in the knowledge that it can either be present, or not. In this way, it has more akin to power than other kinds of resource. (Cue philosophical diversion… I feel a picture coming on… I need a whiteboard… and, back in the room!)
It is also fantastically complex. Deep down, data networks do their very best to exploit the laws of physics: if you see the difference between a theoretical ‘digital’ broadband signal wave and what actually gets sent down the fifty-year-old piece of copper between your house and the box down the road, you start to appreciate the challenge. “How far can we push this thing” is a familiar concept for low-level network engineers.
At higher levels, protocols for data transfers, for streaming, for signalling and one-off events, vary according to both the type of information being transferred and the equipment being used. There’s some fascinating work taking place around ultra-low bandwidth networks, which can support minimal needs (“your cow has a temperature”) with equally low demands on resources.
In other words, networking constitutes a series of trade-offs, trying to get the maximum out whilst minimising risk of failure. We like to think of data networks as single streams of light, but in reality they are like million-lane highway interchanges, each of which has to manage tractors alongside juggernauts, motorbikes with skateboards.
Upon which sandy basis, we build our castle-in-the-cloud existences. As with the webinar yesterday, it was interesting to note the final advice: that any internet-centric approach needs to consider connectivity above anything else. Basically, if you can’t connect, you’re stuffed: this should be so obvious that it doesn’t need to be said, but in reality, precisely the opposite is true. We take our networks for granted, and this needs to be called out.
Smart Shift: We are all makers now
Speaking of resources we could not do without, where would we be without the lowly screw? In this chapter we think about manufacturing and the rise of Alibaba, the Internet of Things and 3D printing.
Thanks for reading! Jon