OS GIS Conference Day 1

Hello from Minneapolis! I'm in town for Open Source Geospatial '05, the biggest gathering of open source map hackers. I'm sharing a room with Simon St Laurent, intrepid editor of Mapping Hacks and Web Mapping Illustrated. It's a busy room: Tyler (author of Web Mapping Illustrated) crashed here on Wednesday night, and Schuyler and Jo's bikes (they're the authors of Mapping Hacks) are by the TV as I write this.

Yup, the authors of both books are here, and Simon got to see the look on Tyler's face when he saw the book for the first time. That's a great feeling as an editor; it's like watching a baby be born but all the screaming and gore happened months ago.

You can tell a good conference by the density of interesting people, and by that measure this is a great conference. When I arrived, everyone was at demos—a little like the Where Fair (and several of the conference attendees will be at the Where Fair) but without the booze. I found an amazing project, MapBuilder, being demo'd by Michael Adair of Natural Resources Canada. It's a massively Ajax in-browser GIS system: multiple layers, images, all pulled from WFS and WMS servers on the fly. There's no programming required to insert a map into a web page, just an XML configuration file, a div tag, and an onload=... addition to your body tag.

WFS and WMS are everywhere. Unlike in the general software industry, where it seems like web services enabled new applications, in the map industry web services solved problems that people were already struggling with. The proprietary data formats and wire protocols had resulted in a culture of plugins and a data interchange/application interop nightmare. They (in the open source arena particularly) fell upon web services like starved jackals and standard APIs are now the norm rather than the exception. There are still a few holdouts in industry, big companies intent on maintaining their vendor lock-in, but increasing governments (huge consumers of GIS systems) are mandating open protocols (e.g., my own New Zealand just did so).

I had a long yarn with Dirk Willem van Gulik, current President of the Apache Foundation, about this on the bus to the evening event. He might be Apache President by night but by day he's a semantic web GIS hacker. He's been making his living doing integration from multiple data sources, each with their own maintenance and quality cultures, in such a way that sources don't have to change anything to be integratable. He's competing with big companies for government business and winning with the help of open source. In Europe, Dirk said, government workers sometimes get edicts from higher-up saying "thou shalt buy [expensive proprietary product X]". They can ignore these edicts and go with open source because the law says that they must use open standards and the proprietary product does not. I wish everyone had the freedom to ignore those kinds of ridiculous mandates from above!

The dinner event was at Fort Snelling, a military base dating back to 1820. They had the doors open, with hosts in period costume to show us around. Dirk and I wandered through the exhibits of medicine, military life, housing, and so on, periodically pointing each other to interesting trivia from the guide book or wall. The fort is in a beautiful location at the junction of the Minnesota and Mississippi rivers, and I had a brief reverie from the top of the round tower overlooking the junction: imagine the fishing!

Dinner was BBQ ribs. I love this conference already. During dinner, Dirk and I sat beside James Macgill, the developer of the Java geotools package. We had an entertaining conversation about the shortcomings of new programmers. Dirk's position was that new programmers don't know the ground truth of programming—assembly language, memory management, threads and other innards of the operating system. As a result you get layers of abstraction through which harsh reality still leaks, leading to subtle errors. Another programmer at the table, Jody Garnett, had a great response: "don't panic, games programmers know all this stuff and they burn out after a few years and enter the pool of open source programmers." I love the idea that rampant burnout in the games industry will help open source.

On the way back, James and I chatted about open source projects. He had to reboot early in geotools' life: he'd been developing it himself, and it was easier to implement things himself than show people how to develop in his code. So he threw it all away and started again: something Joel argues against. It really worked for him, though: they spent a long time designing the new version in a very open process. Anyone who came along and said "your design is crap!" was listened to and their advice usually taken. Now they have great code and a thriving developer community.

One interesting feature of the community around geotools is that that the average lifetime is around 6 months—they get a lot of turnover. They don't panic about this, they embrace it. James encourages multiple programmers on each part of the project, so code doesn't get orphaned when someone leaves.

Day Two began with lightning talks. Tyler, author of Web Mapping Illustrated, was hilarious. I'd drunk with him last night, so I know he's more fun than a barrel of monkeys. The next speaker, though, was Schuyler. He outdid Tyler for entertainment, if that's possible: he proposed a distributed WMS cache to solve the problem that no single person has enough bandwidth to serve imagery to everyone. Free geo data creates a bandwidth problem and Schuyler's about to start a distributed cache project to solve that problem.

My personal favourite, though, was a guy who demo'd OS Earth. He had challenged conference attendees last year to use games technology to put mapping, imagery, etc. into the hands of everyone. Built on OpenGL, it's a desktop app that fetches data from WMS servers seamlessly, lets you zoom in and pan around smoothly. It's beautiful and I can't wait to put it on my kids' computers.