Thursday, October 18, 2012

State Of The Map US - 2012

Thanks to OpenGeo, last weekend I attended the OpenStreetMap State of the Map (US) conference in Portland, Oregon. It was a great conference to attend.  Below are some of my takeaways.

[Obligatory disclaimer: these are my opinions, not necessarily those of OpenGeo]

It was an interesting change from being at software conferences like FOSS4G.  The conference was mostly about software, but from a reverse perspective. I
nstead of new software ideas in search of applications and data, the perspective was of a huge, freely-available, rapidly growing dataset and how this encourages the development of innovative software to manage, display and analyze it.  One result of this was less of a "tribal" split among the attendees. (A tribal rift that might have appeared is choice of mapping engine, but just about everyone there was in the Mapnik tribe. I had to dig to find GeoServer or MapServer users.)

What was similar to FOSS4G was the notable excitement about transforming established business models - but with different players, organizations which use (and usually pay for) road and transit data, rather than established software vendors. Also similar was the feeling of community, the energy and passion of the people involved, and the sense of being involved in something that is a radically new and more empowering way of building something that is essential in people’s lives.
(See the presentation When Google Maps Gives you Lemons, make Lemonade for a great example of why and how this is happening).

Incidentally, the pre-conference party was at GeoLoqui, who later that weekend announced their acquisition by ESRI. Apparently the big money hadn't started flowing their way yet, since they ran out of beer just after we got there.

Statistics
  • about 225 attendees
  • 3 tracks of presentations over 2 days
  • 3rd SOTM-US.  First one was 2 years ago, with only 40 attendees

The list of talks with links to some presentations.

Comments on Selected Talks

The other thing similar to FOSS4G is that there are many more good talks than one person can possibly take in. Here's notes from ones I did attend:


  • Steve Coast (formerly of Cloudmade, now at Microsoft) gave his 1000th "Founder of OSM" keynote.  He said that the top thing OSM needs to continue to grow is addressing.  He singled out Google Mapmaker as being a clear and present danger to the growth of OSM.
  • Check the great video Address is Approximate, mentioned by Henk Hoff in his keynote
  • Dane Springmeyer and Artem Pavlenko from MapBox presented on new features of Mapnik.  
    • Ability to define raster compositing operations between map layers.  Showed using it to obtain tint bands, as well as some other examples which are maybe more interesting than essential.  
    • Vertex converters, similar to GeometryTransformations in GeoServer.  Uses include simplification (various simplification methods were tested - not clear which are available) and smoothing
    • Format support: CSV, GeoJSON, Python
  • Nathan Kelso and Michal Migurski from MapBox presented on how to prepare OSM and terrain data to make a purty base map.  A lot of work is required, and most of it is slow/complex and thus has to be done offline (the beauty of tiles...).  They have a clever simulated-annealing based labeller with a cool video.
  • AJ Ashton from MapBox talked about OSM data preparation for effective cartography.  It makes the good point that good multi-scale cartography takes a lot of pre-rendering data prep, as well as additional data sources to OSM (such as NaturalEarth for placename priority).  
  • MapBox and the Knight Foundation announced their big grant ($575K) to “improve OSM tools”.  There was a palpable sense of concern in the community attending about what this would mean for the current developer community.  Apparently there are still hard feelings about past experiences with CloudMade and MapQuest.  
  • Tom Macwright from MapBox talked about OSM infrastructure.  I had no idea it was run on such a minimal infrastucture (3 servers in London).  The codebase (known as “RailsPort”) is pretty Rube(y) Goldbergian.  The bus factor is a bit too low for comfort.  The good news is that the data fits in a terabyte, so there are mirrors all over the place.
  • Steve Coast raised an interesting question wondering if it would improve the OSM codebase if it had reusability/portability as an explicit goal.  You might think it would be rare to find other people who need to run a massive crowd-sourced map of the world, but Eric Wolf of USGS says that they run a fork of OSM internally to support their mapping.  He did say that it was difficult to contribute back to the trunk.
  • Nathan Van Der Wilt presented his ArgyleTiles project to create a tiled base map of earth imagery.  A great idea, but one that’s been tried before (where have you gone, OpenAerialMap?).
  • David Turner from OpenPlans discussed using OpenTripPlanner with OSM.  They have highly-tuned route planning rules which work with OSM tags.  OpenTripPlanner is a very interesting and useful project.  They make a fair bit of use of JTS and spatial algorithms (such as linear referencing and concave hulls)
  • Portland TriMet presented on their use of OSM and OpenTripPlanner to provide extremely high-quality routing (much better than Google in many cases, particularly for multi-modal involving human-powered transport).  They did a lot of work to improve OSM data in Portland to make routing work better. They emphasized that this was not that expensive - 4 interns over a few months to improve and QA the entire area.  This is an ongoing project, and is done interatively in conjunction with the OpenTrip Planner, to identify routes which show up as taking longer than expected. TriMet also presented on their extensive use of open data for mapping.
  • There were a couple of talks from TeleNav, a commercial traffic reporting and routing company. They have 30M cllients globally reporting and using traffic data. Sounds like they are switching to use OSM data.  They are putting a lot of work into error detection and data cleaning.  They have created an open alternative to the TeleAtlas/NavTeq TMC spec (which is licensed) call TTL.  This creates a standard set of road segments which is the basis for traffic data collection and reporting.  
  • Martijn Van Exel is a longtime OSM user and advocate, now based in Salt Lake City.  He is doing a lot of work on error detection and reporting in OSM, notable creating the Remap-A-Tron.  He’s using GeoServer (yay!).  I told him about the new heatmap rendering, which might be of interest for visualization
  • Jeff Meyer presented on spatio-temporal applications in education and the humanities.  He made a plea to model and capture temporal attributes in OSM, to preserve its value into the future.
  • Abe Usher had a hilarious presentation on Heatmaps for data visualization.  
  • Ben Standefer from Urban Airship talked about how location-based push to mobile devices is becoming big business.  They are using OSM for things like identifying neighbourhood polygons, and POI (points of interest) polygons (eg arenas).  They’re using the JTS STRtree spatial indexing (yay! Blows me away that it is suitable for hard production use)
  • Alex Barth from MapBox talked about Carmen, an open-source geocoder they are developing.  It seems more like a reverse geocoder - most of what he talked about was how to identify named locations from points.  Nothing was said about the hard parts of address geocoding (such as parsing, error handling, address models, and fuzzy matching)