Some crazy academics have written a very academic paper (PDF) on how you can use the same system to figure out where to put sensors in the water to detect contaminants, and make sure we catch all the good stories in the blogosphere. Oh goody!
Basically, all you have to do is exploit submodularity to develop an efficient algorithm that scales to large problems, achieving near optimal placements, while being 700 times faster than a simple greedy algorithm [we hate greedy algorithms]. Also, you need to derive online bounds on the quality of the placements obtained by any algorithm. Their algorithms and bounds also handle cases here nodes (sensor locations, blogs) have different costs.