Xtech 2006: Tom Loosemore – Treating Digital Broadcast As Just Another API, and other such ruminations

Going to tell us a story about Mr Baird, Mr Moore, and Mr Berners-Lee.

10-15 years ago, Mr Baird ruled the roost, but we know about TV and what makes great TV is great programmes, fabulous stories fabulously told. Mr Moore then came a long and said our chips will get faster, our kit will get smaller, and his corollary, that disks will just keep getting bigger. That was 30 years ago. 15 years ago Mr Berners-lee populated the net, and said the ‘internet is made of people’.

10 years ago, Mr Loosemore started working for Wired in the UK as a journalist before they went bust. One of his jobs was to keep abreast of Moore’s Law, as the editor wanted to do a monthly feature on costs and size of computing equipment. Recently found a spreadsheet from 95 charting ISP costs and it was really expensive. In 95 everything was analogue – TV, satellite, cable.

Then in 98 Mr Murdoch gave away digital set-top boxes, and it cost £2billion, but the market thought he was nuts. News International nearly cost him the business. But he saw that it was an essential move, because it gave him more bandwidth. In the UK in 95 you had 4 maybe 5 channels, but when Murdoch went with his set-top box, you had hundreds.

Then digital terrestrial started, which was rubbish, but then taken over by the BBC and you can have about 30 free channels.

Doesn’t look at digital broadcasting the same way that everyone else does. Sees it as a way of distributing 1s and 0s. Doesn’t see it as programmes, but as data.

Lots of different standards and formats.

Also live P2P being used to stream live TV.

Focus on Freeview, and view it as an API.

Expect from an API:

– rich, i.e. interesting. 30 TV channels and a bunch of radio is rich.

– open. Freeview is an unencrypted

– well structured. in theory Freeview is

– scalable

– very high availability, it doesn’t fall over

– accessible

– proven

Doesn’t do so well:

– licence? licence is domestic and personal, so do what you will so long as it is domestic and personal.

– documented? Theoretically, dtg.org.uk. But the documentation is copyrighted and managed by Digital Television Group, so have to be a member before you can get the documentation.

But it’s not hard to reverse engineer, so you can see where the broadcasters are adhering to the standards and where they are being a bit naughty.

Five years ago, Freeview is just taking off in the UK, but other stuff also going on.

There’s a lot of data, 2mpbs MPEG2, 2GB storage per day, 50gb per channel per day, so a terabyte will store 4 channels for a week. But linear TV is a bad way to distribute stuff – most of the time you miss most of the stuff. So what if we just record everything.

So, colleagues built a box to store entire broadcast from the BBC for a week. 2.3 terabytes of storage. About 1000 programmes. Had it for about three weeks. When you’ve got that much choice, existing TV interfaces like the grid layout don’t work. Too much data.

Broadcast metadata alongside the programmes, and the BBC have created an API for that metadata, Got 18 months worth of programme metadata, and got Phil Gifford turned it into a website. Got genre data, but that’s pretty useless when you have 100,000 programmes and it’s not help finding stuff you like.

But if you show people stuff that people are in, say programmes with Caroline Quentin, that’s helpful. Mood data was about as useful as genre, but associate it with people it becomes interesting.

Then discovered the BBC Programme Catalogue. Wonderfully well structured data model, and amazing how disciplined they had been in keeping their vocabularies consistent. So Matt Biddulph put it online, and the crucial thing is that everything is a feed – RDF, FOAF etc.

But that’s only the metadata. Where are the programmes?

So, 12 TB stores all BBC TV for 6 months, and that’s a lot of programmes. But what happens when you give people that amount of content? Can’t make it public, but can make it available to BBC staff, who have to watch a lot of TV in order to do their job. Built an internal pilot, the Archive Testbed, which is no longer live. Took the learning from the metadata only prototype and found a few things.

Keep the channel data. Channels are useful and throwing them away too soon cost them. Channel brands are more than just a navigational paradigm, they are a kite mark of different types of programme. So some programmes scream ‘BBC 2’, for example.

Give people all the metadata, all of which came from external broadcast sources, not internal databases.

Added ratings and comments, links to blog posts, bit of social scheduling – what are my friends watching? What do people recommend? If I don’t know what I want, I want other people to tell me.

Was fantastic, but had to limit it to a couple of hundred people within the BBC. Was a bit too popular for their own good.

In the R&D department, a couple of them worked on a project called Kamaelia to create framework to plug together components for network applications and about six months ago, persuaded them they needed a project for that framework and so applied it to this.

Hopefully will make the project very successful. Now BBC Macro has been released as a pilot. Will be eventually everywhere.

Comments are closed.