I am going to use this post to address Brad's arguments.
On Afghanistan's geographical problems: I think you are right that I somewhat disregarded the geographical problems in Afghanistan. Then again, states can adapt. The British got burned badly in South Africa as the Boers moved to guerrilla warfare and managed to use the environment to beat the British. By the end, the British adapted pretty nicely. Boers were quelled with high civilian losses. We might not like the methods, but it worked.
Same thing with many other civil wars that I have studied. The only thing that kept them going were arms supplies and outside interference. Counterfactually, had the Taliban and AQ not be able to retire to their safe havens in Pakistan, they would have been crushed in the first or second years of America's occupation of Afghanistan.
There is an interesting essay by McChrystal in Foreign Policy, describing the evolution of tactics that the U.S. solders have used to fight against al-Qaeda. It is actually not a new tactic: guerrillas all over the world operate like that, though what really novel is the integrated army-wide counter-guerrilla operation. I may be wrong, but I think McChrystal's strategy might work pretty well regardless of the geographical condition, as long as the Pakistanis plug their side of the border, which is politically and militarily difficult to do.
On Afghanistan's fragmented society: I agree that it is very difficult to build a nation from a patchwork of tribes in Afghanistan. I may oversimplify the problem, but I think the biggest problem in Afghanistan is the rotten central government, which hurts the nation-building capability. A leader cannot exhort others to join the government if they don't see any benefits in joining, and with the central government corrupt, this already difficult task becomes harder to achieve. Get a strong, clean and efficient central government that can do the things that it promised will do, then this problem will be solved.
On the Taliban: My problem with Karzai's trying to reach agreement with Taliban is in his position: whether he is negotiating from position of strength or weaknesses. The Taliban are not idiots. They watch news and probably assume that America's will to stay there is steadily declining. It is true that see a stalemate there, but it is a stalemate thanks to U.S. difficulties in rooting out the Taliban and the Talibans' impossibility to dislodge the U.S. Once the U.S. is gone, the Taliban could simply roll to Kabul. That's a bad politics.
I may sound like a broken radio here, but again, Karzai needs to get this government going. The fact that he is propped up by the U.S. signals weakness. The Taliban may just wait, until the U.S. leaves, the to heck with that scrap of paper called treaty.
On Pakistan: I think the biggest problem with U.S. policy to Pakistan is a lack of alternatives and imagination. The U.S. simply didn't pay much attention to problems within Pakistan - Bush just listened to Musharraf without paying much attention to troubles within Pakistan. Same thing with Obama: Pakistan is a neglected 100-ton gorilla in South Asia, who is deadly afraid at India's 800-ton gorilla next door. On the one hand, the U.S. needs Pakistan because it lacks alternative players that can provides a supply route into Afghanistan, and at the same time, it needs an Islamic state to provide some sort of legitimacy to American actions. On the other hand, the U.S. relies on Pakistan so much that Pakistan knows that the U.S. cannot accomplish much without them.
I think India is the key here. The problem is that Pakistan will not do much unless they can get Kashmir, and India will not talk if the issue of Kashmir is on the plate. India's unwillingness to talk about Kashmir was one of the reasons why Pakistan armed radicals and the Taliban: so they can be used to sow trouble in Kashmir, forcing India to talk.
It is a tough nut to crack, and I think the US really needs to start looking at the region as a whole if we want to solve the Afghanistan problem.