GraphHopper Routing Engine 0.7 Released
graphhopper.com
graphhopper.com
What I still dream of is support for public transport/timetable routing (GTFS). That would be great.
My impression is that GraphHopper is fast and does not take much memory whereas OTP feels quite "heavy" on comparable datasets (city-area level OSM data). So my hope is that if GraphHopper would support timetable routing with the same quality, it will probably beat OTP.
My ultimate goal is to have open-source based multimodal routing on the coutry level (Germany) which would include public transport as well. Frankly, importing country-level datasets in OTP seems unrealistic. GraphHopper on the contrary seems very promising.
http://wiki.openstreetmap.org/wiki/List_of_OSM-based_service...
https://iliauk.com/2016/02/23/millions-of-distances-osrm-pyt...
I never really liked the waze method of manual input, but that's a really braindead way of getting started; it seems however that you could do tricks like anomaly detection of speed of a car through an area when compared to other cars on that street/streets around it with some very hand wavy heuristics. Contributors would install the app and it'd start building a database of avg speed and variance for roads/groups of roads. Anonimization comes to mind as a concern, but since you just need to know "someone drove on that street" at some point recently I imagine there are some things you can do to reduce potential impact/aggregate, since traffic jam data gets less useful in places with less people, although that does require a large starting population, and at this point I'm rambling.
The other advantage of a system like this is that it could readily be used for road discovery, frankly as someone who knows nothing about the open mapping community I always figured something like that _had_ to exist to make it more feasible.
http://improve-osm.org/#37.7251079,-121.6035461,9/layer=OSM/...
It seems to have stalled out, but OpenTraffic was going to build something like you sketch out: