geobloggers

Stefan over at Ogle Earth has just reminded me that I’m speaking at the O’Reilly Web 2.0 Expo this March (Upcoming event here). Should be fun, I’m hoping for some tea and biscuits in the bar afterwards.

Anyway here’s the blurb about my bit that I didn’t write while on medication but probably sounds like it …

—–

Sure geotagging is easy, it’s all latitude, longitude, geoRSS, KML, maps, and what-not. But what do you do when you have well over 50 million geotagged objects? How do you actually do anything constructive with that? With emphasis on the How and constructive parts.

To start with we’ll take a quick(ish) look at the current state of reverse geo-coding; mapping latitude and longitude to an actual place. And what to do when that place is the wrong place or technically the right place but not what anyone calls it. Why it seems as though it should be simple but in reality it’s all terribly hard and we’re still just at the very start of that one. Geocoding != Maps.

With that out of the way, Dan Catt will briefly talk about the database structure at Flickr and how this effects the way we store all the location data for millions upon millions of objects. Then watch in amazement as a fractal curve from 1891 is introduced and marvel at the how it can help us optimize spatial searches over 100 years later.

We’ll then look at the original implementation of the Places project and why it had to be done as offline tasks, what we stored, how and why, and then what we did with it.

After the entertaining elegance of the Places pages we’ll move on to our research into applying clustering strategies based on the fractal curve mentioned above, to pull all sorts of fancy tricks such as working out the geo-clustering of tags and using those as a basis for finding stuff for the Places pages in real-time—turning our 100,000 pre-crunched Places pages into Any Place, Any Time.

Looking beyond the Places page and back to the map, discussing how we are attempting to show the distribution of millions of objects on the map: how you have to pay attention to backend clustering limitations, exactly how much data you can get away with sending to a client’s browser, and the interface and user interactions you need to consider before presenting it all.

Then it’ll all be wrapped up with a look into the future of how we can cut up the Time and Space Cake in various ways to make it easily digestible. An intentionally ambiguous (and terrible) metaphor allowing the sneaking in of any last minute fun and cool stuff.