ASYLUM

* Pickins

Finally posted the last part of the essay series below. Not entirely pleased with it, it didn't so much end as peter out. I think I lost track of where I was going due to being away from it — I was out or busy most nights last week due to a mixture of work and personal stuff. Check del.icio.us for the mini links of fun, but here's a trio of random geeky articles shamelessly stolen from Boing Boing:

• Cable companies are getting screwed for all the reasons I've just been discussing.

• Cuban biotech has taken off bigstyle. They even have 26 patents registered in the USA. Wonder how the State Department feels about that ?

• This Starbucks-based note on threading amuses me, but won't load right now; there's an excerpt here.

* Open-Ended

Technology tends to be a story of increasing specialisation on the part of people inventing things, and increasing generalisation on the part of the things they invent. The two are perhaps related; there's just so much knowledge in the world, that people either need to focus on a small part of it, or else keep a broad overview and hide away the details — we abstract the complexity out of it so that we don't have to deal with all the details.

We don't usually think about the physics of electricity when we flick on a light switch, nor about the OSI 7-layer protocol stack (oy) when we send an email. This is a good thing — I don't want to think about OSI 7-layer ever.

One example of this is the Internet Protocol Stack. This takes care of the business of getting bits of data from A to B for us, a whole messy realm of detail that would crush attempts to get any networking done if we had to worry about it all the time. People can re-use it for pretty much anything, where previously (and in numerous ill-conceived examples since), communications systems would invent their own special language for each damn device. This is stupid for several reasons:

• lots of effort reinventing the wheel each time

• frequently the reinvention is worse than the Internet Protocol

• these devices can only talk to each other, not anything else

• these protocols are often application-specific

While the first two are bad enough — and endemic in the software industry — the last two are the real killers, and what I want to focus on in this post.

Communications devices are useful in proportion to the number of devices/people who can communicate. Alexander Graham Bell's very first telephone was pretty damn useless, because the only other person he could talk to was his assistant, Watson. It's all about the interconnects.

But there's more to it than just how many people your telephone, fax machine, radio, computer or sub-etha sensomatic thumb can talk to: it's about what you can do once you're connected.

Telephones are designed for one thing only: conveying speech. This is pretty useful, so much so that no-one really looked at what else could be done for decades. Not only that, but at first, no-one else was _allowed_ to, because the telephone networks were owned by monopolies who wouldn't let anyone else play. Barriers to entry. No innovation. No progress.

Now, some cunning people came up with a way of signalling computer data across the telephone. This was pretty clever, and got us a long, long way, but it was always a kludge and slow and made inefficient use of the wire. These days many of us have moved on. See, voice can be represented as bits — this is one of those abstractions. We can turn all sorts of things into bits (and then back again), so by inventing a single device for transporting bits, we can reuse that device for voices, text, music, photos, movies, whatever.

These days, we tend to call that device 'broadband' or — for those of us who've been using broadband so long we forget there are people still on dialup — simply 'the internet'.

Of course, for those who follow these things, I'm just restating the end to end principle here, but between Vonage and their ilk, and newcomers like Skype, telephony across the internet has become mainstream, rather than a dismal enthusiast-only mess as it was. As a result, it makes sense to get a fat internet connection and run voice across a small part of that, leaving the rest free for other uses — instead of having a voice-specific connection you can squeeze the internet across if you don't mind being unable to make or receive voice-calls.

Bellheads might blather about Quality of Service and IP latency, but if you've made an international call lately, you'll know that bellhead QoS isn't that hot these days — they're too busy pouring your call into the narrowest space possible to fit more traffic into their existing lines. Almost certainly using VoIP technology, ironically enough.

Bluetooth, 3G

I mentioned in an earlier article, Bluetooth isn't as successful as WiFi. This, despite that it's cheaper and can be used in more 'consumer' devices like phones and headsets, whereas WiFi is generally used embedded into an expensive laptop PC. Not only that, but if you've used it, it's fiddly and annoying.

It's a classic example of bellhead thinking. Bluetooth is made up of a bunch of different 'profiles' — different ways of using Bluetooth each tailored to a particular application, such as wireless headsets or syncing data to your PC. Often, phones won't let you use more than one at once, or only certain combinations. Devices support profiles to varying degrees. Internet Protocol is just one option, so by the time you get around to, say, browsing the web, you've got a bunch of extra layers to pass through.

And, of course, it's owned by a special-interest group of mobile phone manufacturers, who control it, instead of being free for anyone to dick around with.

As for 3G... meh, I'm running out of energy to argue this with. You could check out this BBC article, and remember that when they say "80% of the population", that means a handful of major cities...

The Future

The future is the wireless internet; nothing else much makes sense. The Internet Protocol, once considered kinda complex, is actually trivial by today's standards. You can abuse it a little, cut it down to size based on certain limitations, and put it in a damn light-bulb.

There's an RJ45 ethernet socket out there with its own embedded 486 CPU you can program to do whatever the hell you like, up to and including play Doom. So it's damn near "free" to make an internet-capable device instead of one tailored to a specific technology.

So it makes sense that all radio technologies gradually just become different ways — with different speeds, ranges, battery power requirements — of being part of the internet. There are significant hurdles to overcome, but that's where it's all going, and economies of scale will pound the rest into the dirt sooner or later.

And now I'm going to get a snack and post some nice, short, interesting bits and bobs instead of this giant waffly essay thing.