This is ridiculous in the year 2012. Nothing about this should be a headache (apart from people keeping their data secret for various reasons) yet it's still bafflingly hard.
Suppose that data source A has a point for “Logan Airport” and data source B has a point (a few hundred feet away) for “General Edward Lawrence Logan International Airport” and data source C has a separate data point for every terminal.... on a case-by-case basis it’s not hard to resolve questions like this, but when you try to come up with an algorithm that scales to tens of millions of map features, you spend a lot of time scratching your head saying “why did it do THAT?!?!”