Is It Possible Defeat a Guerrilla Insurgency?

I have come to conclude that the Iraq and Afghan conflicts are virtually unwinnable. Now that we're almost a decade into the conflcits, one has to wonder if a popular insurgency can ever be defeated. Robert Kaplan has a piece in The Atlantic on how the Sri Lankan government defeated the Tamil Tigers. It wasn't pretty, and the US and NATO troops will never be permitted to try the brutal tactics. Here's two examples:

The insurgents are using human shields? No problem. Just keep killing the innocent bystanders until you get to the fighters themselves. There is no comparison between the few civilians that have been killed by American Predator drones in the Afghanistan-Pakistan border region, and the many that were killed by the Sri Lankan government. The Americans have carefully targeted select al-Qaeda members and, in the process, killed a few—at the most, dozens—of civilians among whom the fighters were surrounded. By contrast, the Sri Lankan military indiscriminately killed large numbers of civilians—as many as 20,000 in the final months of fighting, according to the United Nations.

Bad media coverage is hurting morale and giving succor to the enemy? Just kill the journalists. That's what the Sri Lankan authorities did. Precisely because insurgencies are unconventional, there are no easy-to-follow infantry advances and retreats, so the media holds the power to shape a narrative for the public. Aware of the need for a compliant media to aid the war effort, the Sri Lankan government struck fear into the ranks of journalists. There were hundreds of disappearances of top opinion leaders.

In the run up to the Afghan War and more particularly, the Iraq War, was anybody asking if the Americans had a solution for a guerilla uprising? They lost Vietnam that way, you would think somebody might have. I guess nobody thought of it.

It should also be noted that it took Sri Lanka 26 years to totally defeat them. I can't see a democratic society accepting a commitment for that long.