At his much-lauded (and deservedly so) AGU lecture on Successful Predictions (of global warming, a brief history) the delightfully witty Ray Pierrehumbert was asked about the feasibility of geo-engineering. His answer deserves quoting in full, in a text-searchable form:
I see lots of [geo-engineering ideas] that are feasible, but they all terrify me. (Except for schemes for taking CO2 out of the atmosphere, which some people refer to as geo-engineering. Those I find relatively benign, because they put the climate back in the state it was in before we started messing with it.) The feasible geo-engineering schemes that scare me are the crazy ideas to make artificial volcanoes and put sulphate aerosols into the stratosphere.
The reason I think they are barking mad is that you have to assume that we will continue influencing the climate for 10,000 years. You have to renew the aerosols every two years or so. So you're assuming that somehow society will stay together for the next 10,000 years to jam up these aerosols longer than there have been human civilizations, practically. And if you ever stop, than the aerosols go away in a couple of years and you're hit with the full force of global warming…unfortunately, I think the sulphate aerosol injection schemes are probably economically feasible. You don't have to inject too much up there, but it puts the world in a state I call Damocles World. It's like [living in a world forever under] the Sword of Damocles.