Sumantra Bose first hit my radar screen when he wrote the best analysis I have read of post-war Bosnia. Here he combines that research with one other case that I know nearly as well - Cyprus - and three others about which I know much less - indeed I was astonished to realise just how little I knew about Kashmir. Also, although he doesn't give it separate treatment, Northern Ireland is a constant point of reference throughout the book.
I found this a very clear-headed analysis. These are all awful cases of human misery caused by other humans, and great powers meddling irresponsibly (one point he doesn't make, but which struck me, is that the Kashmir and Israel/Palestine situations share the experience of an indecently precipitate British withdrawal in 1947/1948). For all that, there has not been a lot of cross-referencing between them by scholars or practitioners.
The two cases I am more familiar with both essentially have their solutions mapped out - actively in the case of Dayton and Bosnia, potentially in the case of the Annan Plan and Cyprus. Bose does not hesitate to be prescriptive in the other three cases, where a settlement is not currently on the table - the Tamils will not get independence, but must get autonomy, with guarantees for the non-Tamil minorities; there will be no referendum in Kashmir, and the Line of Control will become the permanent boundary, but India has to deliver on autonomy for the area it controls and India and Pakistan must open up the LOC; there must be a Palestinian state, and Hamas must be brought into the political process. He makes the cases compellingly, though my libertarian heart regrets that the Kashmiris will clearly not get the independence that they apparently actually want.
Bose draws two lessons from the five cases. First, that constructive third-party engagement is essential to help move local actors away from zero-sum games. I couldn't agree more. The dog that doesn't bark here, in a way, is Northern Ireland: the 1998 settlement was essentially what was on the table in 1973 (as Seamus Mallon said, "Sunningdale for slow learners"). But it did require an externally appointed chairman of the calibre of George Mitchell to get everyone to agree to what in the end they knew they would have to agree to. Even then, of course, it took another nine years to nail down properly, but (whatever the DUP may say) 1998 is the moment of departure.
Bose's second point is that it is much better to start by aiming for the big picture rather than an incremental approach. This is slightly more controversial, but my instinct is again that he is basically right. The poster child of failure here is the Oslo process in the Middle East, but I've heard it said in the Cyprus context especially as well: in the absence of a big picture agreement (or even the framework of one) within which to operate, negotiating confidence-building measures can be a huge diversion of energy and can actually result in worse rather than better relations between the parties. (Supporters of incrementalism may complain that it was never seriously tried in Cyprus, and never seriously implemented in the Middle East, but perhaps those difficulties illustrate the basic problem.)
One conceptual point which Bose hints at, and I wish he had explored more, is the issue of democracy. In polarised situations, it is almost natural for politicians to try and compete with each other in chauvinism rather than in their willingness to accommodate - Sri Lanka and Israel/Palestine are particularly obvious examples, as indeed is Northern Ireland. This creates difficulties for international peace-builders who (and this is my analysis, not Bose's) will instinctively try to construct "moderates" who are worth engaging with and "hardliners" who are not, essentially judging the standing of the local actors by the extent to which they are prepared to talk pretty for the internationals. Of course, the only criterion for credibility in the end is the level of your popular support; and while it is reasonable to set certain hurdles to participation in formal dialogue, it is stupid to set them in such a way that you prevent the critical mass necessary to consolidate the process from forming. Democracy is a hugely complicating factor in conflict resolution, but also a very necessary one.