Breadth-first search is exponential in time and space, that really won`t cut it on any map of decent size I`m afraid.
Dijkstra`s shortest path algorithm may work well, but as you said, someone needs to implement it, and we need the data (nodes, vertices and ideally weights). Both seem to be lacking at the moment for most countries (for the Netherlands we might be able to do something using using OSM data though).
Bookmarks