In the run-up to the Nato intervention in the Kosovo conflict, I was told by several Nato advisers: "The only question is whether Milosevic will give in just before or just after the start of air strikes." It is easy to be wise after the event, but this phrase illuminates the multiple errors which led Nato into war. To begin with, in the whole lexicon of international relations, there is no such thing as "air strikes." Even a limited armed attack on another country constitutes an act of war. And in war, the enemy can be expected to hit back with every means at his disposal. Furthermore, a war over Kosovo was never going to be with "Milosevic." This war is with the Yugoslav state and the Serbian nation; like so many wars, it began between the Serbs and the Kosovo Albanians over control of a particular territory, and Nato has now ended up on the side of the Albanians. As in any war, a Nato victory will require a partial or even complete Serbian defeat.

The question now facing Nato leaders is the extent of the defeat they can or wish to inflict on the Serbs; what this comes down to, in the end, is how Kosovo is going to be divided or whether Nato means to give the whole province to the Albanians, leading to a voluntary or forced exodus of the Serb minority.

In other words, this means the terms of an ethnic partition. No significant number of Albanians will be able to live safely under Serbian rule in the future-and the much smaller, but deeply-rooted, Serbian population will also not be able to live under Albanian rule. This has become obvious in recent weeks, but the breakdown of ethnic relations in the province was evident in the 1980s; it has roots in territorial conflict going back 1,300 years, exacerbated by the Albanian role in the crushing of Serb revolts against the Ottomans and the atrocities committed by both sides in the wars since the Ottoman empire's collapse.

War is a school of realism, a solvent of established beliefs and an impetus to harsh, but clear decisions. The fact that, in going to war, Nato has stepped outside the usual bounds of international legality (such as it is), should help us to take a hard look at some of the shibboleths on which western policy towards ethnic conflicts have been founded. Having gone to war to prevent the violent suppression of an ethnic rebellion, and having-as will surely be the case-gone on to divide up a state, it would be strange now to return to a rigid adherence to the principles of territorial integrity. The fact that our servicemen are risking their lives in a conflict in which Britain's interests are hardly at stake, should make us focus on what history can tell us about achieving lasting ethnic settlements.

in a recent essay on Kosovo in the New York Review of Books, Timothy Garton Ash argued that "good fences make good neighbours"; contrary to current western beliefs in "integration" and "multiculturalism," the best chance for the Yugoslav peoples to progress would be as separate nation-states with clear ethnic majorities. This has, after all, been the pattern across much of western Europe over the past few centuries, and the process is not finished yet.

This argument goes back to liberal nationalists such as Giuseppe Mazzini, ideologist of the Italian risorgimento. He and his contemporaries argued that real cooperation between European nations could only come after those nations had separated themselves into free and democratic, but also independent, national states. The whole project of European integration can be said to be founded on this argument.

The main objection to this argument is that ethnic nations are often mixed up together and cannot be separated without bloodshed (witness the conflicts over the Italians in South Slav and German lands). And given the disasters which have befallen Europe in the 20th century, many historians mourn Mazzini's great enemy: the multi-ethnic Habsburg empire.

But the Habsburg or Austro-Hungarian empire-like the Soviet Union in our own time-did not in the end survive. Moreover, the Habsburg empire possessed virtues denied to most multi-ethnic states: the state was founded neither on nationality, nor on popular sovereignty, but on supra-national principles: dynastic legitimacy; the rule of law; and the Catholic Church. Further, the Austrian half of the empire contained a number of large nationalities, none of which was in a position completely to dominate the others-unlike the Kingdom of Hungary (where the Hungarians dominated absolutely) or other "multi-ethnic" states, where one dominant majority faces a single large, restive minority. Finally, the empire possessed a state language, German, which in those pre-1914 days was a cosmopolitan language of European civilisation capable of assimilating people of many ethnic origins-as witnessed by the flowering of Jewish thought and culture in Habsburg Vienna.

Elsewhere, by contrast, states which began as multi-ethnic were able to assimilate minority peoples into the dominant nationality by means of the power and prestige of the dominant language and culture-at least if they began the process long ago in pre-modern times. Thus France achieved its 20th century ethno-linguistic homogeneity (until the arrival of the new immigrants) by means of a mainly peaceful, but culturally ruthless, process of destroying the traditions of minorities such as the Bretons, above all through the national school system. Russia would have done the same to Ukrainians and others if it had been able to start a bit sooner. It is hopeless, however, to expect that linguistic cultures such as Serbian or Albanian, the reflection of one small ethnos on the periphery of European civilisation, will ever be able to play this role in assimilating large ethnic minorities-especially where these have access to the culture and the mass media of neighbouring ethnic homelands.

Tito was called "the last Habsburg," and Yugoslav communism (like Soviet communism) was an attempt to overcome national divisions by a supranational ideology focused on the worship of a quasi-emperor. But although Yugoslavia was not (until today) threatened from outside in the manner of the Habsburgs, its internal cultural resources as a multi-ethnic state were weaker. With the waning of communism, some sort of Serb-Albanian war over Kosovo was likely. Several such wars have occurred over the past 100 years, and the struggle in fact resumed with the Kosovo Albanian protests of 1981.

The way in which Nato and its political masters misunderstood the real dynamic of events in Kosovo reflects a characteristic failure of the liberal mind (including most of the western left and the great majority of Americans), which clings to a basically optimistic view of human nature. Such a habit of mind finds it hard to grasp that certain nations really are implacably at odds over the control of ethnically-mixed territory. Instead, the automatic belief is that the innocent masses have been led astray by evil individuals (Milosevic)-or, for the left, by evil ruling classes. These in turn are not motivated by emotions of nationalism, pride or hate, but by "rational" ones of the defence of their political or economic privileges.

This illusion was of little practical moment during the cold war; but thanks to the decision to turn Nato from a defensive alliance into a force for democracy, stability and even market economics in Europe, it has become quite dangerous. Nato has been trapped into becoming the instrument for the policing of ethnic conflicts and the administration of ethnically-mixed societies through protectorates. This is what the Dayton accords have created in Bosnia; it was what the Rambouillet accords laid down for Kosovo; and such a protectorate is also taking shape in Macedonia.

The precedent for this development is the League of Nations "mandate" system, introduced at the Congress of Versailles for territories in the middle east and Africa. This was usually a barely-veiled formula for indirect imperial rule by one of the victorious allies, which is how most of the world sees Nato's aggression ("US imperialism") in Serbia. And, to play the mandate role effectively, we will need something akin to imperial qualities, albeit of a civilised kind.

When divided nations cannot be ruled, mandate-style, by a moderately impartial force, the only "solutions" will be very illiberal ones: either victory by one side, leading to the subjugation or flight of the other's population; or partition and population exchange; or some combination of the two. Although it is natural to recoil from such outcomes, the unwillingness of western diplomats even to admit their possibility in several parts of the world is strange. Not merely has the "international community" accepted several such partitions and population moves in the course of this century, this is exactly what we have ended up with in Bosnia. Between 1992 and 1995, an unwillingness to accept partition helped to prolong the war. Today, it has committed us to the hopeless task of trying to turn Bosnia back into a working multi-ethnic state (assuming that it ever was such a state). An international role in policing Bosnia is essential-but those who pretend that this is leading to re-integration are deceiving themselves and everybody else.

The only realistic choice for Macedonia may also soon be a Bosnia-type partition or a Nato protectorate. Discussions of a possible division between the Albanian, Serb and Macedonian Slav areas are reportedly already taking place in private between Albanian representatives and pro-Bulgarian Macedonian Slav "nationalists"-the idea being that most of Macedonia would join Bulgaria. Even an agreed partition of Macedonia would be a dangerous development, but it may be even more dangerous to keep an African-style artificial legacy of empire in the middle of the Balkans. Macedonia might be able to form a stable multi- ethnic state if its national proportions were stable-but they are not. In fact, the Albanians, who a generation ago were about 20 per cent of the population, may now be as much as 35 per cent thanks to their higher birth rate. By 2020, they could become a majority. It is doubtful, however, whether the Slavs will simply go along with becoming an increasingly small minority in what would ultimately become a de facto part of a Greater Albanian confederation.

Partition and population exchanges are certainly not a universal solution. But when everything else has failed, we need to have the courage to take responsibility for the solution. The alternative is to summon up the ruthlessness to support one side in the conflict-as in Croatia-or be prepared to commit ourselves to the long-term, heavy and sustained policing role necessary to prevent two hostile and mingled populations from tearing each other apart.

in the end, policing and security are what these conflicts are about-the actual physical security of individuals at risk. This raises an important point about the nature of modern nationalism. We must distinguish between classical nationalism and ethnic (or one might say skinhead) hooliganism, and the role of such hooliganism in ethnically-mixed societies. In other words, it is often not a question of protecting a whole population from assault by state-backed ethnic forces, but of protecting families and individuals from being attacked by members of other ethnic groups with the acquiescence of the police-a problem with which we are not wholly unfamiliar in Britain.

A Russian who had moved to Russia in 1992 from a village in southern Kazakhstan once explained to me how this kind of thing occurs: "There were only four Russian families in our village, and many of the Kazakhs wanted us out. People started jeering at me on the street, saying-'we're going to fuck your daughter,' that kind of thing. Nothing happened, and perhaps nothing would have happened. I'd lived there for years, and I had good friends among the Kazakhs. But we knew that the police would never have done anything if we were attacked. So we got out."

Recent decades have indeed seen a decline in certain aspects of classical nationalism, both in the west, and, to a lesser extent, elsewhere. This has been associated with the demilitarisation of society and the spread of modernisation. As a result, there has been a reduction in the willingness to fight for disputed national territories. But this has been paralleled in western Europe and elsewhere by the persistence-or even growth of-ethnic hooliganism. I was in Germany in 1992 at the height of the attacks on immigrants by skinhead and neo-Nazi groups. Headlines outside Germany raised the spectre of the return of aggressive nationalism in Germany, but such fears were almost totally misplaced. These skinheads were up to slouching a block or two in order to terrorise a helpless Turkish family or kick to death an African asylum-seeker; they were not about to join the army, risk death and spend years in trenches to reconquer Danzig for a new German Reich. Their motivations were local: a desire to defend working class jobs, maintain pure neighbourhoods and support local football teams. They did not flow into wider national conflicts because the minorities being targeted have no territorial conflict with Germany and are too small in number to have a significant political presence.

But as we know from Northern Ireland, it is quite different in an ethnically-mixed society where conflicting national-territorial claims exist. There, "skinhead" nationalism and classical nationalism run into each other. Arson, robbery, beatings, rape or murder against members of the other community may be communally encouraged, excused and sheltered. These will contribute in turn to mobilisation along defensive ethnic lines, and to the reliance of ordinary people on any force-however hateful-which will protect them, their families and homes. Such violence does not even need to be condoned by the national leadership of the country or the community concerned. In Chechnya, for example, neither under President Dudayev nor under President Maskhadov has there been any official policy of driving out the local Russians. Dudayev's wife is Russian; and Maskhadov-like Dudayev, a former Soviet officer-always seemed sincere in saying that he felt no hostility to the Russians as a people. None the less, the position of those Russians in Chechnya who enjoyed no protection from a powerful Chechen individual became increasingly desperate; as a result, most had left even before the Russian invasion of December 1994. Chechen traditions of banditry and armed violence are constrained when it comes to attacks on other Chechens, but are unconstrained when it comes to non-Chechens, infidels and especially Russians. A sharp contrast may be drawn with Latvia and Estonia, which have limited the rights of the Russian immigrants, but by peaceful and legal means-with the result that there has been no violence in return.

Where a tradition of socially-sanctioned violence and banditry is present in both parts of an ethnically-mixed society, the only way of containing this in a reasonably equitable way is sustained policing by an outside power-something of which the British empire had immense experience. But for this to work, a whole series of factors have to be present, which Nato can barely begin thinking about in the Balkans. First, the outside power has to be in control of local law and order, and capable of handing out real rewards and punishments. It needs to have a sufficient number of its own men in command of forces on the ground, speaking a local language and understanding local society. In this context, the unarmed EU policemen in Bosnia look quite fatuous.

Second, the presence of the outside power has to appear pretty much eternal. As the British experienced in India in 1945-47 (and on other occasions elsewhere), when the population at large, and locally recruited officials, have a strong sense that the outside power's rule is provisional, then no order will be implemented which risks compromising the official concerned with his own community. No police officer who values his future career or perhaps even his life is going to pursue a charge of ethnically-based theft or murder against a member of his own group. I saw this process at work among Lithuanian and Georgian KGB officers in the Soviet Union in 1990 and 1991. They were still afraid of Moscow, and they would have gone on obeying Moscow if they had been sure that the Soviet Union would last. But because its survival was questionable, it was safer for them to pretend to obey, while in fact doing discreet deals with the nationalists. This contributed significantly to the crumbling of Soviet control over those republics.

Third, the outside policing force will have to adopt an approach very different to that common in modern western societies. In circumstances where criminality is ethnicised and enjoys communal protection, it may be more effective and less provocative to resort to a system of collective punishment; not, of course, mass executions, burning of villages or deportations, but fines, confiscations and restrictions on movement. All of these were employed by the British in India.

If this seems rather strong to some readers, it suggests that they should not choose a career in the policing of ethnic disputes. Because the question of security for individuals and communities is what any settlement of such conflicts is about. The inability of Nato to provide such security in Bosnia has been responsible for the curious state of semi-partition now operating there. On the one hand, Nato and the EU have not been able to accept the truth of Bosnia's three-way partition and withdraw; both because of shibboleths concerning territorial integrity, but also because while the Croat and Serb regions would survive, the Muslims, like the Palestinians, would be left with an unviable two-part state. On the other hand, Nato is incapable of providing the kind of policing on the ground which would allow refugees to return home (Muslims to Serb areas and so on) and begin to make a reality of the reintegration of a united Bosnia which Nato has as its formal objective. Instead, Nato has to work by attempting to coerce Serbian, Croat and Muslim authorities-without the means to control developments at village level.

Policing a violent, ethnically-mixed area is difficult enough even before atrocities have occurred. Afterwards, it is impossible. This is what makes so many peace plans seem like so much waste paper. In the southern Caucasus at present, western diplomats are giving their sympathetic consideration to Georgian suggestions of a peace settlement for the rebel territory of Abkhazia which involves Abkhaz acceptance of Georgian sovereignty, including the disbandment of the Abkhaz army, in return for various constitutional guarantees, including chairmanship of the upper house of the Georgian parliament. The Abkhaz war, parts of which I witnessed, was small-scale, but very brutal. Georgian commanders threatened to kill or expel every single Abkhaz; the Abkhaz ended the war by winning and expelling almost every single Georgian and killing a large number of them. Many of the atrocities on both sides were committed by neighbours who knew each other (also true of Bosnia). Do western diplomats expect that in these circumstances the two populations will agree to live together again? Is it rational to think that any Abkhaz leader will resign the safety of his people into the hands of the Georgians-or that a Georgian would do so if the case were reversed? For my own part, if I knew that someone in my town had killed my parents, and there were no other way of getting justice, sooner or later I would kill him-that is the human response. It is different, of course, if, as in South Africa or the Soviet Union, the atrocities have been committed not by neighbours but by the servants of an anonymous state, with whom the relatives of the victims are not required to live in close proximity.

The Abkhaz case also illustrates how changing demography can affect such disputes. In some cases, like Abkhazia, the Baltic States, Fiji or Malaysia, this has been because of immigration, but elsewhere because of the tendency of certain ethnic groups to have more babies-sometimes as part of a conscious strategy of outnumbering the ethnic enemy. Thus the Albanian proportion of the population of Kosovo rose from 68.5 per cent in 1948 to 77.4 per cent in 1981 and more than 90 per cent before the ethnic cleansing. In Bosnia, the Muslim proportion rose from 31 per cent in 1948 to 43.7 per cent in 1991, while the Serbs fell from 44 per cent to 31 per cent. In both cases, the nations concerned resorted to arguments about which community had been there first and had originally been the largest. In these circumstances, trying to solve ethnic disputes by majoritarian democracy at a given moment is inadequate.

Americans find such dilemmas hard to understand, because they live in an immigrant society, where they have been accustomed to seeing the ethnic composition of whole regions transformed in a single generation. But, in America's case, the immigrants presented no threat of ethnically-based territorial secession. "Native Americans" could have presented a threat, but they were few in number and were soon disposed of, by Serbian methods.

In Britain, however, we have a more persistent and bitter experience. In 1926, catholics in Ulster made up 33.5 per cent of the population (down from 40 per cent 65 years before), and protestants 62.2 per cent-with the catholics basing their national claim on the principle that up to the late 16th century the native Gaels made up 100 per cent of the population. Thanks to their higher birth rate, by 1991, catholics were 38.4 per cent of the population and protestants only 50.6 per cent (the remainder of the population refused to state any religious allegiance). But if the catholic birth rate in Ulster had been a bit higher, or catholic emigration a bit lower, it is possible to imagine a situation in which by 1975, say, the catholics would have been a majority, and could have voted democratically for union with the Republic, which the British government would presumably have felt bound to accept. But does anyone imagine that the main protestant groups would have accepted "the will of the democratic majority"?

In these circumstances-as in Ulster now-the only solution appears to be some form of guaranteed power-sharing with permanently fixed ethnic proportions. This is now very popular in political science; it is called "consociational democracy." But such systems are inimical to many aspects of democracy, because they limit the ability of the electorate to bring about any real change in government-they are more like a medieval estates system. Moreover, they are almost always held in place by an outside force: in Ulster, by the British and Irish governments and the British army; or, in the first modern example of this kind, Moravia, by the Austrian army. The Soviet Union operated a similar system in the ethnically-mixed republics of the north Caucasus-now dissolving in the absence of the Soviet army.

Where such systems have broken down, atrocities have often resulted; if no new imperial rule is on hand, the least worst solution is partition. But one official of a UN agency dealing with ethnic conflict told me: "Officially, we're not even allowed to think about partition." Western diplomats intone "territorial integrity" like a prayer. In Africa, the notion that colonial borders are inviolable has been elevated to a sacred principle. But almost every Africa expert would agree that it is unlikely that in the long run-say, in 200 years' time-the borders established by colonialism, will remain. To point out that the "long run" begins today, that several states have already split along ethnic and historic faultlines, and that it would be better to recognise the fact and work with the results, is taboo.

in the course of this century, the west has accepted or even promoted partition and population exchange on a number of occasions. In 1923, Greece and Turkey negotiated a bilateral agreement whereby some 1.2m Greeks left Turkey for Greece and about 500,000 Turks moved in the opposite direction. In 1948-49 in Palestine, the UN brokered a partition agreement, which was rejected by the Palestinians. During and after the ensuing war, some 750,000 Palestinians were driven out. The west has in effect accepted this, on the grounds that their return would destabilise the Israeli state. UN resolutions on the subject have therefore remained a dead letter.

The deportation of Germans from central Europe (and of Poles from Ukraine, Belarus and Lithuania) in 1945-46 was a ghastly affair (to say nothing of the Holocaust); yet the resulting ethnic homogeneity has allowed the states concerned to become more politically stable than they were between 1919 and 1939, and than their surviving multi-ethnic equivalents have been in the 1990s. A Poland with the ethnic proportions of 1939 would be racked with violent Polish chauvinism and ethnic secessionist movements.

The most infamous case of partition and population exchange is India and Pakistan in 1947, in the course of which at least 300,000 people died. Many of these deaths could have been avoided if the exchange had been properly supervised. Congress and the Muslim League promised that the rights of minorities would be respected after partition. But the violence was largely an upsurge of hatred and fear from below, by populations maddened by years of propaganda.

The division of India and Pakistan has proved an enduring wound for the subcontinent, burdening it with hatreds, wars and grotesque levels of military spending. All the same, it is hard to imagine that a united India would have been a better prospect. Would a constitutional compromise have lasted? The minimum demand of even moderate Muslims was for a loose federation and for some form of permanent guaranteed power-sharing at the centre. This would have led to a desperately weak and unstable state. Under the pressure of Islamist radicalism, Muslim population growth and Hindu nationalism, such a state would very probably have disintegrated. Such a disintegration would have been more chaotic, and perhaps even bloodier than the events of 1947. It might have led to the separation of still more states from India, and the descent of the region to west African levels of disorder and poverty. As it is, India has remained a rather successful constitutional federation, while its Muslim minority, although distrusted and sometimes attacked, is too small to provoke an overwhelming wave of Hindu fear and anger.

To mention India in 1947 is to underline how partition and population exchange should be the last resort in any ethnic dispute. But it also underlines how those outsiders who oppose such a solution by force or diplomatic pressure have a responsibility to prevent future conflict. The British by 1947 were no longer capable of keeping the peace in India, and so were right to get out. Nato must be prepared to police parts of the Balkans permanently, or it must bring about settlements which, however harsh, will prevent future instability. From this point of view, the Rambouillet accords were deficient in one key respect. They set up an elaborate system of "consociational democracy," full of checks and balances, and they provided a Nato force to guarantee these. But they did not guarantee that the Nato force would stay.

The partition of Kosovo-let alone Macedonia-and the acceptance of solutions involving partition for the Caucasus will take courage on the part of western leaders, both because it would expose them to howls of protest from different sides and because it would require deals (especially with Russia) which will be unpopular with the US Congress. But at a time when western servicemen are being required to show physical courage in the field, it seems not inappropriate to ask their leaders to show both moral courage and historical awareness.