I'm getting ready to travel home now from GUADEC 2004, the main annual gathering of GNOME developers and users. It's been a fantastic time, and I've enjoyed getting to know lots of the GNOME hackers better. I gave two talks this year, on GNOME Bluetooth and Metadata for the Desktop.

This latter topic proved to be particularly timely, the talk coinciding with Apple's announcement of OS X Tiger, which includes a metadata-based search very similar to that that demonstrated by Nat Friedman on Monday, Beagle. As Tiger also contains a feature called "Dashboard", there were not a few puzzled frowns and grumpy harrumphs around the conference.

In my talk I made the case for the importance of a metadata framework in the desktop platform. In one sense it didn't need making, as the trend all around is to make metadata-based systems to aid findability on the desktop. I think however it is important to understand why we need metadata, and that integration is as much a goal as findability.

I then proceeded to explain what RDF was and give a brief introduction to the Redland toolkit. My intention was to provide enough information for people to start experimenting. I suspect some people would rather I had presented a ten-point plan for implementing my bright idea, rather than me just urging them to take a look at the technology.

This all set me thinking about innovation in GNOME, and what was necessary to get it going. It seems inevitable that the moments of true breakthrough have come from dedicated individuals or teams with vision and often money. Nautilus, OpenOffice, Evolution. You can't sprinkle fairy dust on the grassroots hackers and expect something marvellous to emerge. The army of GNOME maintainers does a wonderful and fantastic job on the desktop but they're unlikely to be the source of anything but incremental change.

I wish Nat had been able to follow through strongly on Dashboard 18 months ago when the idea first decloaked. Although Dashboard may not have flown, he'd have gotten to the metadata-based search point pretty quickly, and today GNOME might have been a step ahead of OS X. (There are plenty of other questions here which I'll save for some other time, such as whether GNOME should aim to be first or best, or serve consumer as well as enterprise users.)

As I said in my talk, Nat and Miguel have an uncanny history of getting things right. I think we have to be more daring and gamble on their instincts, and those of people like them. We need some way in GNOME to support emerging applications as well as the desktop release. While the focus is the desktop release, the reaction to new ideas will always be questioning whether they're appropriate for the direction of the desktop. While honourable in intent, this inevitably leads to negativity.

I'm going home energized and excited about GNOME and the Linux desktop, with a million plans of what to do. But, as ever, with painfully limited resource.