Microsoft helping OpenStreetMap

However... my experience from the GPS industry is that users quickly become frustrated with even minor faults in map data. It leads to the routing engine giving you back weird routes, like routes that wander through multiple parking lots, zigzag around through residential areas, or just take streets that are always busy. We had a lot of trouble with TeleAtlas' map data (their US data is terrible).

The worst part is that the people writing the routing algorithms start to put in fudges to "clean up" the map data (scare quotes intentional). These are tested by looking at a few areas on the map that the routing engineer knows well. What he doesn't see is that his fudge usually makes things worse in a lot of other parts of the map. The right solution is to fix the data.

It's hard to crowdsource map data. One of the big GPS vendors-- I think it was Magellan-- rolled out a system that let ordinary users edit the maps on their devices. They also rolled out a way that users could contribute the data back very easily. Unfortunately, what they found is that the biggest use people found for the map editing system was to insert fake blockages on roads they didn't want the routing engine to take! People often prefer to do things the quick way rather than the correct way.

Unfortunately, unlike code, it's hard to audit map data contributions from the community. Linus can check a kernel patch to see if it make sense. But if someone from Podunk, East Texas tells you that there are 5 stop signs on Main Street, how do you evaluate that? Nobody back at HQ has ever been to Podunk. We had this problem all the time when working with map data. Unfortunately, it was usually resolved by shrugging and saying, "well nobody lives there anyway."

The silliest part about all of this is that the data for where the roads are, the signage, speed limits, and so forth is usually known to someone in the local government. It's just that the municipalities don't do a good job of putting that data in a standard format and making it available. Well, ok, TIGER data is available, but it's of poor quality-- much worse than TeleAtlas or NAVTEQ. Hopefully these guys will get more accurate, and better at coordinating with the community, in the future.