This is the code that defines my "best location" strategy. http://code.google.com/p/foursquared/source/browse/main/src/com/joelapenna/foursquared/location/BestLocationListener.java
I basically have to deal with the fact that I might get location-from-gps after I get location-from-wifi-tower OR i might get location-from-wifi-tower after I get location-from-gps so that the Last Known location may in some cases return a more or less accurate location than what the phone is really capable of reporting just because I wanted to be able to get a quick-kind-of-accurate-lock while I was trying to accquire a accurate-as-possible sort of lock.
How would I make this better?
I could see an API where instead of registering with individual location providers, you would query the system for a set of location updates. You would define that you want FINE location, but will accept COARSE updates first. You would have a getBestKnownLocation instead of a getLastKnownLocation that would return a location based on heuristics you might define when registering your location provider. Visually, the updates you'd receive would look something like the growing/shrinking circles you get in the gmaps app when honing in on a location. Maybe the developer would request a "HyperLocalWhileWalkingAround" strategy that could be swapped with "PassiveUpdatesWhenLocationIsSignificentlyChanged." both of which could have knobs for controlling specifics of the strategy.
In short, I want to abstract away the code I have shown as an example above. As a end-developer, I don't care where the location comes from, just that I get a location but I do care, what the data looks like.