I've finally got around to reading Ahmed Rashid's Taliban and Descent into Chaos. They are as good as everyone says. Specifically, there are perhaps three things that set Rashid apart as a writer on Central Asia. (His contacts book is outstanding, but then, he's not the only one.)
First of all, he writes about Central Asia, rather than about American politics as expressed through the foreign-policy establishment. He writes about Central Asia in the sense that he places the complex regional politics, the competition for power among Pakistan, Iran, and Russia, at the centre of the story. In fact, you could make a case that the Taliban as a phenomenon is almost irrelevant; if it didn't exist, and the political situation was otherwise unchanged, something else would be playing the role of Durrani Pashtun caucus, drugs logistical system, sink for Saudi malcontents, and Pakistani proxy.
He also writes about Central Asia in the sense that he emphasises the intelligent agency of the regional powers, the Afghans of all allegiances, the Pakistani political parties, and the intelligence agencies. This implies taking a very calm view of the actual extent to which the Americans ever controlled anything in the area - in fact, one of his key points is that the current chaos is largely due to the absence of a US policy in the 1990s, and in many ways, its continued absence up to 2005 and even now.
Rashid also provides an interesting view of Taliban sociology - his version of them is essentially another of the child-soldier and refugee camp movements of the 90s, strongly mutually similar across an arc of suffering from Afghanistan to Liberia. He suggests that their ideology is more of a substitute for Islam and for Afghan culture than anything else - a cut-down set of tropes for people brutally removed from the real things.
Of course, there are certain functions that any good tyranny needs to fulfil. You have to have outward signs, so that it is possible to enforce conformity and identify a hated target-group; it may actually be better if they are content-free, so as not to limit flexibility. You have to have exemplary violence, and again, it may help if it isn't actually directed towards victory. Eliminating people for no reason is the ultimate costly signal that anyone could be a target. No tyranny can function without denunciation - arguably, it's more important than all the other functions. And you need the possibility of competitive observance, in order to get individual initiative on your side. "Working towards the Führer" is the classic example. This may actually be a more useful view of the Talibs of the 90s - a sort of minimal dictatorship.
Finally, he provides an integrated systems view of the politics, economics, and societies involved. The creation of the Taliban, in his view, involved many overlaid political networks, those of the Pakistani trucking industry and its partners in organised crime, those of the Sindhi feudal landlords who were frequently investors in the trucking and smuggling business and also powers in the PPP, the Saudi-financed system through which international jihadis were recruited, fetched to the training camps, supplied, and sent out as cannon fodder to pursue Pakistani aspirations in central Asia, and the ISI.
From a purely Anglo-British point of view, it's worth noting that he is very hard on the Americans about the intelligence picture available to the NATO powers in 2005 when Rumsfeld finally dropped his opposition to ISAF deploying outside Kabul. He strongly supports the line that, having maintained practically no presence there and diverted their satellite and other reconnaissance resources to Iraq, the Americans let 16th Air Assault Brigade deploy into a zone of the unknown, which in the way of these things turned out to be full of the enemy. If true, this is the second occasion on which they've welshed on the agreement under which the UK doesn't operate its own imagery satellites. Rashid argues that there was a vital window of opportunity to get a broad-based political settlement in 2002-2003, which the Cheney administration* squandered in the interests of invading Iraq and pleasing the ISI/Saudi intelligence services.
In general, I can't escape the conclusion that Kashmir is still the issue.
(* - Rashid's view of the last US administration is very much a Cheney administration)