Over on CNet, Nate Lanxon has delivered a stonking piece on the technical background to the BBC's iPlayer servers.
Which does raise the question, though: is it value for money?
I'm trying to work out how much more - or less - it costs the BBC to deliver a TV programme over iPlayer than it does to transmit it in the old-fashioned way.
This is surprisingly difficult to do, but I think I have an answer: it costs around sixteen thousand times as much to stream a programme as to broadcast it.
Here's how I came to that conclusion. At gigabit speeds, Internet bandwidth costs around a penny per megabit per hour (I'm assuming the BBC is buying it in at around £7/Mbps/month, and there are 744 hours in a month). An HD stream is around a megabit, so the cost to the BBC of one person watching one programme for one hour is a penny. More or less.
Now, the BBC 1 analogue TV transmitter at Crystal Palace is a nominal megawatt. Analogue transmitters are very inefficient, so it could need twice as much actual power - if it was actually transmitting a megawatt. Because of the structure of a TV signal, most of the time it's transmitting far less. And I don't know the cost to the BBC of the electricity (and anyway, the BBC doesn't operate the transmitters itself any more). So with all those variables, I'm going to just assume it's taking a megawatt, and that a megawatt/hour costs around £100.
Crystal Palace analogue TV has a catchment area of around 15 million people. Despite the BBC's very best efforts, they don't all watch BBC TV all the time - the average is 20 hours a week, or about eight percent of the time. Which is 300 million person-hours per week, or 1.8 million person-hours per hour - and that hour costs £100. That means one person watching one hour of TV from Crystal Palace costs the BBC sixty micropence.
That's sixteen thousand times cheaper than streaming.
Any better guesses?