Once talks begin, they will not be easy. Nato will probably have little choice but to stick with the Rambouillet formula on autonomy but, in practice, Kosovo can never again be more than a notional part of Serbia. The scale of post-war reconstruction and the provision of security for returning refugees will mean that it must be in effect a Nato protectorate, whatever face-saving form of words may be found.
The main feature of any settlement must be that Western troops are based in Kosovo. This is the single point on which Nato cannot equivocate. The only serious solution to the mounting refugee crises in Macedonia, Montenegro and Albania is for people to return to what is left of their homes, but they cannot, dare not, venture back until their security is guaranteed against Serb troops and militias. If the alliance is unable to achieve this, then it will have been defeated.
This was the issue on which it went to war in the first place and on which the Rambouillet talks foundered. President Slobodan Milosevic rejected all entreaties to accept an armed international force in the province to reassure the people; it was essential to his strategy that they be insecure. So far, despite Nato claims of cracks opening up in the Yugoslav government, there has been absolutely no sign of a shift in its position. Once the concession is made, however it is dressed up politically or diluted by the inclusion of Russian and Ukrainian troops, Serbia has been defeated. Serbs in Kosovo will also be uncomfortably aware of the grim revelations likely to emerge once their behaviour can be properly scrutinised.
The only possible compromise would be effective partition of Kosovo. If this is proposed by Belgrade, some in Nato might be tempted to accept. But it would be impossible to get Kosovar agreement to such a move. This would create a continuing source of tension, with Serbia fearing Kosovar attempts to seize back any land taken from them. So we must therefore be clear that the main purpose of diplomacy will be to ease the pain and humiliation of defeat for one side.
Victory probably will come to Nato, but it will have come at a heavy price, the result of a flawed strategy that allowed Milosevic to seize the initiative at the start of the campaign. As Nato's air strikes began last month, the Yugoslav decisions they were designed to influence had already been taken and the plans they were supposed to frustrate were well advanced. Success now for Nato requires reversing the tragedy it failed to prevent. It has ruled out the only sure way of achieving this - a land invasion to wrest the territory from Serb control. This is because of a fear of heavy casualties, the treachery of the terrain and the lack of a decent transport network making it difficult to get heavy armour to the front. Although Nato's land power in the area is slowly being reinforced, largely to deal with the refugee crisis, it would take weeks to assemble and train a proper invasion force. Even the introduction of US ground-attack helicopters has been delayed because of the primitive state of Albanian facilities.
So Nato has been left waging war against Yugoslavia's military infrastructure in the hope of persuading Milosevic to admit defeat. The targets in Serbia are described as being military in nature, and they are certainly chosen - and advertised in advance - to avoid excessive casualties, but their impact is also civilian. Nato can continue to work its way through the Yugoslav air force and barracks, but in this bizarre modern warfare they are not immediately relevant to the problem at hand. The attacks on fuel, factories and bridges are another matter. They are steadily ruining the already shaky Yugoslav economy. This is undoubtedly putting pressure on Milosevic to find an escape route, but there are no obvious manifestations of anti-war feeling. If anything, nationalist sentiment has hardened. As the inadvertent air attack on a train on Monday illustrated, there are limits to the number of civil/military targets that can be struck without harming civilians. Although Nato's current tactics do leave many people uncomfortable, the Serbs are paying a heavy political price for the success of their Kosovo campaign. The tide of human misery it created hardened European opinion against Yugoslavia and created a level of commitment within the Nato nations that will be difficult to undermine. It is too late for Nato to conduct the air war over Kosovo to impede "ethnic cleansing", but it can create the conditions for the refugees' return. This requires depleting the fuel and ammunition stocks of the local Serb units, isolating them from Serbia so that they cannot be reinforced or resupplied, disrupting their command and control networks, obstructing their attempts to move by knocking down bridges and keeping them vulnerable to attack by constant daylight air patrols. This may release Nato from its dependence upon Milosevic's consent by returning refugees with an armed Nato guard. There are here at least the components of a plan that could bring this war to a conclusion on Nato's terms.
When this is coupled with the economic costs of the war, the outlook for Milosevic is therefore grim. His regime appears to be clinging to the hope that somehow, through threats or aid, Russia can extract them from the mess, but they have taken statements from Moscow far too literally. Milosevic himself is experienced enough to know just how little is on offer. The most useful role Russia could now play would be to dispel any wishful thinking in Belgrade (thereby also making themselves more acceptable to Nato in any future attempt to broker a deal).
The failure to prepare for a land option has left Nato over-dependent upon coercion to bring this war to an early conclusion. This can work only by persuading Belgrade that the costs of further resistance are much higher than the costs of compliance with Nato's core demands. There is absolutely no way of predicting when this conclusion will be reached or if it will ever be reached at all. It is, however, most likely to come when Milosevic is convinced that Nato forces will eventually enter Kosovo without his consent.
The author is professor of war studies at King's College, London