Syria’s civil war seems unstoppable. In four years, it has killed 250,000, displaced half the country’s pre-war population of 23 million, and given birth to the Islamic State, which promises to conquer the region and inspire lone-wolf terrorist attacks throughout the Western world.

The Obama administration has managed to keep the United States from a third major conflict in the broader Middle East, but its Syria policy has otherwise struggled. President Bashar al-Assad did not, as expected, tumble from power as his money ran out and his army collapsed. Efforts to create a moderate military opposition have generally failed. The Geneva peace process, intended to fashion a new coalition government, has gone nowhere. The new American-Turkish plan to create a safe zone in northern Syria may not amount to much, given the lack of available forces to establish and protect it. And even with a settlement, who will keep the peace in a country where the Syrian army has so much blood on its hands, the Islamic State and the al-Qaeda affiliate known as al-Nusra are so extreme in their ideology, and the moderate opposition is so weak and fractured?

To envision how the Syrian civil war might ultimately end, it’s helpful to think through how civil wars end in general, and then ask which approach might apply in this case.

The most obvious way to end a war is through outright military victory — when one side defeats the other, as when the Rwandan Patriotic Front rose up and responded to the Hutu-led genocide in 1994. Scholars such as Harrison Wagner argue that this is the most stable of various possible civil war outcomes. Of course, permanent victories are often elusive, as defeated groups lick their wounds, regroup and plot to fight another day. And in Syria, the two strongest groups – Assad with his army, the Islamic State with its jihadists – are both unacceptable winners for Western interests.

A second potential war-ender is an intervention by some outside power, which could side with one party to win it. Beyond the U.S.-led wars of the last 15 years, good modern examples include Tanzania overthrowing Idi Amin in Uganda and the Vietnamese army defeating the Khmer Rouge in Cambodia. But only Turkey (or Israel) has the means to do this in the region, and only the United States has the capacity to mount such an operation from afar. Both are nonstarters, since neither Ankara nor Washington (nor Israel) is interested in such a role.

Another option is a negotiated compromise that brings peace. This is basically the approach being attempted in Syria today. But it tends to work only when parties to a conflict are exhausted, typically by a decade or more of fighting (as with Angola and Mozambique after the Cold War or parts of Central America), or when one side recognizes that it cannot win the war but has not yet been completely beaten (as possibly with the FARC in Colombia now). Moreover, it requires some trusted peacekeeping force or relatively trusted army to implement the peace deal — as scholars such as Barbara Walter at the University of California/San Diego and Paige Fortna at Columbia have analyzed. Syria has no such neutral military force today.

The final option is partition or confederation. Partition is certainly easier said than done — whether the goal is to create new countries or autonomous zones held together through some weak central government. But if the parties do recognize that they need to work together and there is some natural way to divide up land that is seen as both fair and militarily enforceable, partition can work. Conflicts between Bosnia and Kosovo, between Eritrea and Ethiopia and between the two Sudans have all ended this way — though often only after a great deal of blood has been spilled, and often only with the help of international peacekeepers along the various lines of separation.

Of these four, only the last seems realistic for Syria. Even this plan would be hard, given how many of the country’s central cities are ethnically intermixed. But it has a chance, whereas other strategies offer little hope.

The core reason is this: Any Syria deal is going to have to be enforced. And at present, for the first three options listed above, there is no viable candidate for enforcing it. The parties to the conflict are too extremist, too mistrustful of each other and too far from any clean military outcome. No foreign power has the will and means to impose a solution countrywide.

But envisioning a federal arrangement offers the hope that a future peacekeeping force in Syria that could deploy largely along the lines of separation rather than throughout all the major populated areas. That would reduce its needed size and its likely casualty levels. There would surely be violence, and tests of the force — so Americans would have to be part of it, to give it backbone and credibility, but to the tune of perhaps 10,000 to 20,000 troops rather than the 100,000 or more that typified our peak efforts in both Iraq and Afghanistan. Moreover, even this kind of deal would require the defeat or near-defeat of both the Islamic State and Assad, given how divisive and illegitimate each has become. So it would only be possible after moderate opposition forces had been strengthened and made much more military headway than they have so far.

This points a path forward. The United States and partners should expand their help for moderate factions, among other things by relaxing the vetting standards that have prevented us from working with anyone who wants to target Assad rather than just the Islamic State. Once somewhat larger moderate forces are available, and able to establish dependable toeholds within Syria, we should send in training teams to work with them in accelerating the recruiting and training of local forces. Such an approach would also allow the much better provisioning of humanitarian relief — an urgent priority recognized by all.

This confederated approach has one more advantage that the other paths do not: It achieves some important goals even on an interim basis, and even if it fails to achieve its preferred aims. By creating gradually expanding “ink spots” within Syria based on various moderate groups, we can help get food and medicine and schooling to millions who are currently deprived, while also creating new outposts from which to monitor and at times attack ISIL.

A slow-motion genocide is happening in Syria, and it is also putting American allies in the region as well as American lives around the world at risk. We do not now have a viable strategy for addressing it. Working toward a confederal Syria provides the best hope of finding a serious way ahead.