Thanks for all the great work on this. The street list is getting better over time. I’d like to offer a slightly different approach. The community approach is important and we need to have it, but there’s still a lot that can be done safely with automated rules.
Now, we know OpenStreetMap is not perfect and can have bad data. But some elements just simply cannot be confused with runnable features and can safely be removed by algorithm.
If we can set a few filtering rules that we can all agree on, it would eliminate 90 percent of the non-runnable features (at least in the cities that I frequent).
Here’s my proposal:
-
Eliminate any feature with the tag building=yes. There is no possible chance that a feature rendered as a building on OSM should be treated as a runnable feature.
-
Eliminate any feature with a waterway key, regardless of value. Lakes, rivers, streams, etc., are not runnable and cannot be confused with roads or paths.
-
Eliminate all leisure=park features. In fact, you might consider eliminating anything with the leisure key with the possible exception of leisure=track. I have found that the leisure=park feature is used very commonly and is rendered as a park boundary area.
-
Eliminate any feature with a landuse key. These are also rendered as boundaries and will never be runnable features.
By excluding obvious non-runnable features we can reduce effort but still take a conservative approach that won’t inappropriately remove streets and paths.