Kaplan has a good summary of the intelligent debate about how to approach Afghanistan (CT is counter-terrorism, this approach is said to be favored by Joe Biden) ...
... Some in the CT camp realize that the COIN-dinistas (as critics call them) have a point. Their real gripe with counterinsurgency is that it costs too much and promises too little. Even most COIN strategists acknowledge that a successful campaign, especially in Afghanistan, would require lots of troops (way more than President Obama has committed so far), lots of time (a decade or so), and lots of money (wiping out most or all of the savings achieved by the withdrawal from Iraq)—and even then the insurgents might still win.
A "targeted" CT campaign, its advocates say, would at least demonstrate the West's resolve in the war on terrorism and keep al-Qaida jihadists contained. It's a type of fighting that we know how to do, and its effects are measurable. One might also argue (I don't know if anyone on the inside is doing so) that it could serve as a holding action—a way of keeping Afghanistan from plunging deeper into chaos—while we focus more intently on diplomatic measures to stabilize neighboring Pakistan. If Pakistan blows up, curing Afghanistan of its problems will be irrelevant and, in any case, impossible.
Some in the COIN camp have sympathy for this argument—especially for the part about the high cost and the uncertainty of success—but they would argue back that a purely CT approach is sure to fail in the long run...
I think we ended up going the COIN route in Iraq. It's too early to know how well it worked, but it seems to have been an improvement on than every other approach that was tried there.
One thing we know, however, is that we're financially and militarily exhausted.
So the real debate is between containment, and a Pakistan stabilization strategy, and a big investment in Afghanistan.
I sympathize with everyone, and especially the people of Afghanistan.
The one consolation is that I think we have a good team struggling with the problem.