There is a lot of talk around better algorithms (e.g. contraction hierarchies), distributed routing, stress tests etc. So I'm going to put in into perspective with a few calculations.
For a 40km journey, Gosmore takes 50ms*. So let's say Errol costs around $10,000 and you want to pay it off in 2 years by "selling" routes. Over the two years it can calculate roughly 12 billion 40km routes. That's 0.0001 cents per route. So you can make a profit by selling 2 year subscriptions at 1c each. Compare that with how expensive cars are per km. An algorithm that is 10 times faster does not change the economics. By contrast, let's say you add support for just a few more tags. Perhaps routing the driver around congested intersections. He sees that the product is saving him time and fuel. Then he will pay when you increase your price with a few dollars. *: I admit that the core algorithm is quite bad but it does make up for it by being able to do everything in RAM and reducing cache misses. _______________________________________________ talk mailing list [email protected] http://lists.openstreetmap.org/listinfo/talk
