We are therefore being pushed inexorably in the direction of a ground offensive - probably not by the whole of Nato, but a "coalition of the willing" involving the US and UK. Despite all its risks, a ground war would still be better than an air campaign lasting months and not just because of the effects on Serbia - strange though it may be to mention these under the circumstances.
Months of bombing would leave an economically ruined, embittered Serbia as a source of anti-Western terrorism and destabilisation. For all foreseeable time Serbia would be a "rogue state". The West may now be promising some kind of Marshall plan for the Balkans, but it is highly unlikely that any conceivable postwar Serbian government would qualify, given the furious anti-Western feelings it will represent.
However, the dangers of a ground war are by now well-known. Attacking from Albania involves formidable logistical obstacles. An offensive via Macedonia or Montenegro looks politically impossible at present. Attacking from the north (assuming that Hungary and/or Bosnia were to agree) is militarily much simpler, but would involve the occupation of the whole of Serbia, with everything that would mean for terrorism, partisan warfare and political quagmire.
Before moving towards a ground war, we therefore need to make a serious attempt to reach a compromise - not with Milosevic, who is by now largely irrelevant to the real long-term issue, but with the Serbian nation. It is with this nation that we are now at war, and it is with this nation that we shall still be dealing when this war is over.
The mission of Strobe Talbott, the US Deputy Secretary of State, to Moscow this week gives us the chance to use Russia to achieve such a compromise - and this would also do something to restore our dangerously shattered relations with Moscow. However, for this to work, we cannot simply follow US policy so far, which has essentially been one of telling Russia to tell Belgrade to surrender. If we are going to get the Serbs to agree to a settlement without defeating them on the ground or bombing them into starvation, then any settlement will obviously have to be a negotiated one.
To achieve such a settlement first of all means facing facts. The most important of these is that Albanians and Serbs cannot now live together in a common state - indeed, this was already apparent 10 years ago. A large number of Albanians cannot possibly be asked to live again under the threat of Serbian military terror; but, equally, a withdrawal of Yugoslav forces from the whole of Kosovo and their replacement with Nato troops would sooner or later lead inevitably to the formal separation of the whole of Kosovo and the flight of its Serbian minority. Nato will not stay there indefinitely to protect them.
To achieve such a result, in an area that Serbs regard as of critical historical and emotional importance, will require their complete military defeat. To maintain such a settlement against the threat of a Serbian revanche will require a permanent Nato commitment in the region. It would probably also involve the heavy arming of Albania by Nato - a grotesque proposition, considering Albania's record as a failing state.
The only way to get out of this hole is to partition Kosovo. This could begin with the creation of "safe havens" for the Kosovar Albanians covering some two-thirds of the existing territory of Kosovo and akin to those established for the Kurds of northern Iraq in 1991. Yugoslav forces would withdraw behind a line patrolled by UN forces with a Nato component and guaranteed by the UN (including Russia) as well as Nato. This would allow most (though by no means all) of the refugees and those displaced within Kosovo to return home and would secure their safety. Nato would, however, have to give a formal assurance to Yugoslavia that, unless Yugoslav forces crossed that line, Nato air attacks would not resume.
This solution would obviously mean the Serbs giving up most of Kosovo. Its attraction for them, however, would be that it would involve de facto partition. This would leave them in control both of the remaining areas of Serbian population and of the emotionally most significant places: the battle field of Kosovo Polje itself and the Serbian monasteries.
This would be achieved by leaving in Serbian hands both a large swathe of territory in the north and a small, separate bit in the west adjacent to Montenegro and containing Pec (the site of the old Serbian Orthodox Patriarchate) and the monastery of Decane. This division would give the Serbs a great deal more than their present share of the Kosovar population would justify, but it would be roughly in line with their historical position in recent centuries, before the higher Albanian birthrate radically changed the local balance.
Such a solution would obviously be unacceptable to the Albanians. In particular, leaving Kosovo Polje in Serbian hands means in effect doing the same for the capital, Pristina, of which it is now a southern suburb. However, it should be possible to compensate the Kosovar Albanians for this loss. As it is, a massive programme of aid will be needed to rebuild Albanian homes destroyed by the Serbs. There is no reason why this should not be extended to the construction of a new regional capital at Prizren in the south-east - especially since this was in fact the capital of Kosovo under the Ottoman Empire.
A solution involving safe havens and de facto partition obviously involves Nato giving up some of its existing principles for a settlement, but it could still just about be presented as a Nato victory; but are we prepared to fight for a complete victory and live with the consequences?
The author's `Chechnya: Tombstone of Russian Power' is due out in paperback in JuneReuse content