The BBC is quietly preparing a challenge to Microsoft and other companies jostling to reap revenues from video streams. It is developing code-decode (codec) software called Dirac in an open-source project aimed at providing a royalty-free way to distribute video .

The sums at stake are potentially huge because the software industry insists on payment per viewer, per hour of encoded content. This contrasts with TV technology, for which viewers and broadcasters alike make a one-off royalties payment when they buy their equipment.

Tim Borer, manager of the Dirac project at the BBC's Kingswood Warren R&D lab, pointed out: 'Coding standards for video were always free and open. We have been broadcasting PAL TV in this country for decades. The standard has been available for anyone to use... If the BBC had to pay per hour of coding in PAL we would be in trouble.'

The cost of the Real Networks player used on the BBC site will become prohibitive as the number of users rises, Borer said. 'Ultimately we would like to support millions of users with streaming services. It is uneconomic to use a codec with a per-user licence.'

There is the technical hassle, too. 'You need a back channel from the media player right back to the server to audit how many channels are being played. That complicates the entire system.'

The Dirac codec will infringe no patents and will be freely available worldwide for any platform including Windows, Linux and the Mac OS, potentially undercutting the market for commercial formats. It can be used for passing video round home networks, rights-managed peer-to-peer file sharing, or playing media in handheld devices, as well as for web streaming.

Borer sees open-source as a natural fit with the BBC's not-for-profit, public service remit. 'We have a different ethos to a commercial company,' he said. 'Open-source software development is really a parallel development. It's got the same ethos.'

Controversial player

The media player used on the BBC site has been controversial from the start. The BBC standardised on the Real Media module, opening itself to accusations of favouritism, because it did not want the complication and expense of offering all major commercial players.

Users who downloaded the player from Real Networks also complained of having to negotiate screens apparently trying to trick them into buying the full player, though the system has since become more straightforward.

Dirac began as a project to compress high-definition TV but researchers realised it had wider applications both for internal desktop production and external streaming. The immediate aim is to be able to decode standard digital TV definition (720 x 576 pixels at 25 frames per second) in real time; the current version can decode a quarter-standard resolution at 20fps (frames per second), which is enough for Internet streaming.

'I'd expect the decoding speed to have been sorted out certainly within the next year... We have already done some optimisation work and have had large speed-up - of the order of 10 times,' said Borer. The figures assume a 3GHz processor but the focus is on gaining speed by code optimisation rather than hardware because the BBC wants Dirac to be usable on a broad range of devices.

Bolt-on software libaries

Borer's team is trying to make project more accessible to open-source developers; lead Dirac programmer Anuradha Suraparaju is developing an interface to facilitate use of the module by C coders. 'They can simply bolt on a software library to their existing application,' said Borer, who hopes the developer community will write Dirac plug-ins for players such as Windows Media Player 9.

The Dirac format will be submitted to relevant organisations for ratification as a standard. Usually only a bitstream is standardised, with the code and decode methods being left to the implementers; but the BBC will be providing implementation.

Borer believes Dirac could turn out to be more efficient than standards based on commercial patents, even though it has to use technology more than 20 years old to avoid breaking patents.

Many of the techniques were published by academics long before they were used commercially, and some are relatively new in being applied to video compression. 'Wavelets [exotic waveforms used to map changes] have been around for 20 years now but they have only recently been ratified as part of the JPEG 2000 specification,' said Borer.

Efficiency depends on how techniques are used as much as on the techniques themselves. Commercial organisations trying to agree a standard currently fight to get their own intellectual property included, which makes for complexity and does not necessarily lead to the adoption of the best of breed. 'We have tried to make our codec as simple as we can so it is easy to understand and easy to implement efficiently,' said Borer.

Cost benefits

Unsurprisingly a number of commercial and public organisations have expressed interest in a project that could cut their own costs; and Dirac, like other open-source products, can act at the very least as a restraint on the greed of commercial developers.

Borer says the BBC would be happy to use commercial products, provided they had a viable licensing model, and he insists that he is not trying to compete with them. 'We are happy to work with them,' he said. 'We work a lot with Microsoft. It is purely that we wish to preserve open standards.'

Changes in the air
How BBC archive might be put online. The first part of our special report on the BBC's Kingswood Warren lab.

A different way of watching
How metadata could change the way we view TV. Part 2 of our special report on the Kingswood Warren lab.