Max Fisher wrote recently on Vox that partitioning countries is seldom pleasant. He also reluctantly posited that in Iraq today, it may be the least bad option.
Why the reluctance? Partitioning countries along ethnic or sectarian lines is, historically, the way most viable nation-states have been made. To see why, consider two outcomes from the last time a country broke apart into several: Yugoslavia.
Serbia and Croatia
The Balkan Wars effectively tipped off when Croatia tried to secede, and a majority-Serb region of Croatia -- the Krajina -- counter-seceded. The central question of the ensuing war was: "Will there be a Croatia, and if so, what will its borders be?"
The Croats settled this question by winning. In 1995, they conquered the Krajina and evicted nearly everyone in it. It wasn't pretty, but it was decisive and everyone knew it. Two decades on, the countries live peacefully side by side, with Croatia in the EU and Serbia angling to join. Compare that with this:
Bosnia and Herzegovina
Bosnia was the most polyglot of Yugoslavia's seven states, and the site of its most infamous massacre during the war. Today's it's an internationally-midwifed, multi-ethnic society... but it's also divided and unworkable. Its citizens mostly live in different regions, threaten to secede often, vote along ethnic lines, and have parallel administrative structures. Their government is among the world's most complicated and least productive. A senior Balkan diplomat once told me that "Bosnia was only put on the U.N. Security Council [in 2010-11] to keep it from splitting up."
Bosnia now depends on indefinite international stewardship. That can work for a small country of 3.8 million in a stable neighborhood, but for a sprawling nation of 33 million in the heart of the Middle East, it cannot.
So what does this mean for Iraq?
At this point, a united Iraq would, at best, work like Bosnia does. In other words, it wouldn't. A partitioned Iraq, properly configured, would work like Croatia and Serbia do -- unpleasantly created, but viable and peaceful in the long term. More importantly, nearly all the concerns Fisher raises about partition will be worse if Iraq stays together than if it doesn't.
Partition won't internationalize the conflict
It's not true, as is commonly said, that partitions internationalize a conflict: unfinished partitions do. Pakistan-India, Israel-Palestine, and Sudan-South Sudan never agreed on borders, meaning that unresolved intrastate problems became interstate problems. In Iraq today, "Should there be an Kurdistan?" is therefore less of a threat to peace than "Who gets Kirkuk?"
Croatia demonstrated the most effective way around this problem: create indisputable facts on the ground. This is done by conquering territory and expelling unwanted minorities, and like it or not, the parties are already doing this in Iraq. The longer the outside world waits, the more ruthlessly effective this process will be.
Partition won't worsen life for minorities
In the short term, minorities suffer hugely in partitions, often being forcibly displaced in large numbers. But this is already true in a united Iraq: UNHCR estimates more than 3.5 million Iraqis are internally displaced. Long term, a clear partition removes the existential fear of other groups, encouraging moderation. Five years after the committing the largest act of European ethnic cleansing since World War II, Croatia, now secure in its own borders, invited all displaced Serbs home. (Half returned.) Serbia, meanwhile, overthrew its strongman Slobodan Milosevic, who died in the Hague.
Partition will require less, not more, of America
A united Iraq appears to require a permanent (and unwanted) American presence, but a divided one would not. Washington would need only the capability to repulse an overt attack by one state against another, something it's pretty good at doing.
Partition will discourage Saudi-Iranian power jockeying
A united but weak Iraq almost forces Iran and Saudi Arabia to compete for influence, with ruinous results. By contrast, a clearly defined partition, particularly between the Sunni Arab and Shiite Arab regions, would also yield clearly defined spheres of influence, leaving the regional heavyweights with nothing to fight over.
Partition is happening anyway
With the Kurds increasingly making a play for national liberation, Sunni conquerors breaking on the rocks of Shiite regions and vice versa, an Iraqi (and possibly Syrian) breakup looks more likely by the day. Letting the parties get there on their own is better, both morally and strategically, than helping one side crush the others to save an arbitrarily-created state that has never shown it can be governed well.
The history of partitions is fraught, but an Iraqi future modeled on Croatia and Serbia is viable. A future modeled on Bosnia and Herzegovina is not. Partition is the least bad option. It may be the only good option.