Wednesday, September 26, 2007

Review of FOSS4G

I've been up at the FOSS4G conference in Victoria this week, which has just finished (apart from the related "code sprint" tomorrow). It was really an excellent conference - congratulations to conference chair Paul Ramsey and the rest of the organizing committee for putting on a great event. The quality of the sessions I went to was consistently high, and there was a real energy and buzz around the whole event (much more than at most of the more established geospatial conferences I have been to recently). Adena Schutzberg said in her closing comments that her overall impression of the conference and the open source geospatial community was one of maturity (and she will expand on that theme at Directions magazine next week). The event reaffirmed the belief I had before coming here that the role of open source software in the geospatial industry will continue to grow quickly. And personally I enjoyed learning a lot of new things and meeting a new crowd of people (as well as quite a few old friends).

One specific aim for me in coming here was to learn more about PostGIS, which I regarded beforehand as the front runner for the database technology to use for my new company Spatial Networking. Paul Ramsey, who was a busy man this week, giving a number of very good presentations, presented an interesting set of PostGIS case studies. These included IGN, the French national mapping agency, who maintain a database of 100 million geographic features, with frequent updates from 1800 users, and a fleet management company which stores GPS readings from 100 vehicles 10 times a minute for 8 hours a day, or 480,000 records a day. In a separate presentation, Alejandro Chumaceiro of SIGIS from Venezuela, talked about a similar fleet management application with very high update volumes. Interestingly, they use partitioned tables and create a new partition every hour! Incidentally, I talked with Alejandro afterwards and it turns out that he worked for IBM on their GFIS product from 1986 to 1991, and knew me from those days - it's a small world in the geospatial community :). Kevin Neufeld from Refractions Research also gave a lot of useful hints about partitioning and other performance related topics. Brian Timoney talked about the work he has done using both Google Earth and Virtual Earth on the front end, with PostGIS on the back end doing a variety of spatial queries and reports, including capabilities like buffering, in a way which is very easy to use for people with no specialized knowledge. And Tyler Erickson of Michigan Tech Research Institute talked about some interesting spatio-temporal analysis of environmental data he is doing using PostGIS, GeoServer and Google Earth. Overall I was very impressed with the capabilities and scalability of PostGIS, and was reassured that this is the right approach for us to use at Spatial Networking.

Another topic which featured in several sessions I attended was that of data. As Schuyler Erle said in introducing a session about the OSGeo Public Geodata Committee, a key difference between geospatial software and other open source initiatives is that the software is no use without data, so looking at ways to create and maintain, share, and enable users to find, freely available geodata is also an important element of OSGeo's work, in addition to software. Nick Black, a fellow Brit, gave a good talk about Open Street Map, which is getting a lot of interest. The scope of what they are doing is broader than I had realized, including not just streets but points of interest (pubs are apparently especially well represented!), address information which can be used for geocoding, and they are working on building up a US database based on TIGER data. The ubiquitous Geoff Zeiss, a man without whom no GIS conference would be complete :), gave an interesting review of the wide variety of government policies with regard to geospatial data around the world. One curious snippet from that was that in Malaysia and some other Asian countries, you need to have an application reviewed by the police and the army before being able to receive a government-issued map! In the opening session, I enjoyed the talk by Andrew Turner of Mapufacture on Neogeography Data Collection, which was a great overview of the wide range of approaches being used for "community generated" data, including things like cheap aerial photography using remote control drones from Pict'Earth - they have a nice example of data captured at Burning Man. This was one of a number of five minute lightning talks, which went over pretty well - several people told me that they enjoyed the format. I also gave one of those, on the topic of the past, present and future of the geospatial industry, and managed to fit into the allotted time - though next time I might choose a slightly more focused topic :) ! I will write up my talk in a separate post at some point (it will take a lot longer than 5 minutes to do that though!). Ed McNierny of Topozone had the most intriguing title, "No one wants 3/8 inch drill bits" - the punchline was that what they actually want is 3/8 inch holes, and we should focus on the results that our users need, not the specific tools they use to achieve those. Schuyler Erle gave one of the more entertaining presentations on 7-dimensional matrices that I have seen (and I say that as a mathematician).

Also in the opening session, Damian Conway gave a good talk entitled "geek eye for the suit guy", on how to sell "suits" on the benefits of open source software. Roughly half his arguments applied to geospatial software, and half were more specific to Linux - Adena has done a more detailed writeup.

Brady Forrest of O'Reilly Media gave an interesting presentation on "Trends of the Geo Web". His three main themes were "Maps, Maps Everywhere", "The Web as the GeoIndex", and
"Crowdsourced Data Collection". One interesting site he mentioned that I hadn't come across before was Walk Score, which ranks your home based on how "walkable" the neighborhood is (my loft in downtown Denver rated an excellent 94 out of 100). It seems as though every time I see a presentation like this I discover some interesting new sites, and now I listen slightly nervously hoping that I don't discover someone doing what we plan to do with Spatial Networking, but so far that hasn't happened!

I also was on the closing panel for the conference, which I thought went well - we had a pretty lively discussion. The closing session also included a preview of next year's conference which will be in Cape Town, South Africa. I had the pleasure of spending a few days in Cape Town in 2002, followed by a safari in Botswana which still ranks as the best of the many trips I've done to different parts of the world (check out my pictures). So I certainly hope to make it to the conference, and highly recommend that others try to make it down there and spend some additional time in that part of the world too.

Apologies to those I missed out of this somewhat rambling account, but the Sticky Wicket pub is calling, so I will wrap it up here, for now at least.

2 comments:

Nixta said...

Awesome write up. Lots of stuff for me to get my brain caught up on there (although I might give the 7-dimensional matrices a miss, and I say that as someone who failed miserably at university maths).

Andrew Turner said...

It was a great conference. Interesting to see the integration of more corporations into FOSS.

The slides for my lightning talk "Beyond GPS" are available on HighEarthOrbit and the links are at http://del.icio.us/tag/BeyondGPS.