I've always thought OSM's data license should be more BSD-ish than the current LGPLish license. If you were trying to privately build and edit data from OSM, there'd be a pretty strong incentive to get most of your updates and corrections upstreamed back into OSM, (especially for things that are broadly of interest to the rest of the community) because eventually someone will also create those updates in the upstream OSM, and then you'll have to deal with how to merge your changes with the OSM changes.

I think OSM would benefit more with more people building on the dataset, because even if a smaller percentage actually contributed back, they'd make up for it with a larger overall set of people building on the dataset. And I think they'd get a larger set of people building on the dataset with a more permissive license.

(The obvious counterpoint is OSM is doing just fine with its current license)