It's going to take me some time to fully decompress and write this up, as a good portion of what I saw in Portland was new to me in some form. It is easy enough to read about the hacker community and fiddle with the cutting edge stuff like GeoCouch: it is quite another to actually meet the people and see what they are cooking up.
QTDesigner (Holly Glaser), GeoCouch/GeoJson (Max Ogden)
I hopped between sessions on this one, so I missed about half of Max's talk. The QT talk was small and had David Turner walking Holly through making a survey plugin.
The latter session was showed off an interop tool for converting shapefiles into GeoCouch and why you might want to use that datastore over say PostGIS. The larger point here, which was only briefly mentioned and which I'd like to examine more closely at a later date, is the assertion that simplicity in this sphere breeds adoption. I don't know I necessarily agree.
Domination of the Shapefile (James Fee)
I don't recall the real title, but the alt title was "Why I was Wrong in my last Keynote". It turns out Spatialite, though it was an improvement over the traditional method of bulk data delivery (masses of shapefiles), didn't have enough going for it to actually see mainstream adoption. There just isn't enough wrong with shapefiles to bother switching for data providers. Worse is simply better, at least in this case.
Still, there are problems in data interchange that need to be addressed, the big ones being (1) styling, (2) attribute relationships. I honestly don't see too big a problem with the shapefile being multiple files, since people have a tendency to zip up such things anyway and you probably want more than one shapefile at a time.
This was an interesting discussion and probably merits its own post, the examples were particularly illuminating (FGDC, GML, Gulf Spill map styling).
Keynote (Nathaniel Kelso)
Nathaniel works for the Washington Post making beautiful maps. He showed off examples - the accuracy of the coordinates for WhereCamp, Top Secret America, people estimation at the Obama inauguration - around the theme of "filling in" the map. Of an estimated 2.3 million habitated places we have good maps for a small fraction of them.
Multimodal Trip Planner (David Turner)
This talk is an odd intersection of network topology, real-time scheduling, dynamic mobile resources, and human behavior. When people say they only want to travel on bike lanes, does that mean they'll go five miles out of their way to do it? What if they want to bike only downhill by cleverly leveraging Portland's excellent mass transit system?
One thing I didn't bring up, and wish I had, is that this is an app that essentially lets people plan their travel around a relatively fixed mass transit schedule. What happens when everyone is using such information all the time? Would it be possible to start dynamically modifying the transit system itself to best integrate with user demands?
Geodata for the Masses (various)
I caught part of this talk late after briefly stopping in on the OSM editing and civic apps sessions. By the time I got there it appeared to have devolved into a talk on DIY data collection by kite and balloon.