One of the recurrent complaints Bernard Williams has about most contemporary liberal political theory is, roughly, that it underestimates the extent to which some slightly modified version of Clausewitz's axiom that 'war is politics by other means' is not only true but constitutive of politics. For example, he says of political decisions that they do
not in [themselves] announce that the other party was morally wrong, or indeed wrong at all. What [they] immediately [announce] is that they have lost.
The upshot of this is that he is unwilling to attempt to place any serious limits on the legitimacy of political disagreements as such. Because politics is about conflict over the distribution of the various benefits of social cooperation, as long as such conflict continues to present itself as an argument about the distribution of the various benefits of social cooperation, rather than a naked grab for power on the part of some group or other and hence no longer about social cooperation, that fact of conflict is regulative: placing limits on conflict would be an attempt to prevent the relevant groups from disagreeing about what they can disagree about, and thus an attempt to breach politics' regulative ideal.
Tom Nagel, in his 'Equality and Partiality', has a slightly different argument to the same conclusion. He sees, correctly in my view, the problem of political legitimacy as the problem of
[w]hat, if anything, can we all agree that we should do, given that our motives are not merely impersonal?
However, in his schema, this personal perspective is to be opposed to the impersonal perspective, which is the sole source of value. For Nagel, then, phrasing the problem of political legitimacy in this way means that definitionally non-evaluative considerations have weight in any discourse of political legitimacy. As long as I can be resolutely and stubbornly selfish for long enough, I do not have to adduce any reasons which any one else would regard as conclusive or even weighty in favour of that selfishness: the fact that, so far as anyone can tell, whatever I gain from being selfish really matters to me is enough to show that it ought to be given weight of some sort in designing a set of rules to distribute the benefits of social cooperation.
Clearly, Nagel's problems really begin with his sharp distincton between the impersonal and personal perspectives: the entirely accurate description of holding out for something which can have no reasons given in favour of it as selfish shows, fairly conclusively I would have thought, that evaluative language is ineliminable from a personal perspective, whatever that might be anyway. Given that, it is very odd to think that simply being sufficiently stubborn serves to legitimate any demand that someone could make: the question of that demand's evaluation is already an open one, and so why we should abstain from making judgements on it is fair from clear.
Williams seems to have exactly the same problem. Indeed, given his deep commitment to the idea that thick evaluative concepts are an intrinsic and unavoidable part of normal discourse, an idea he thinks the importance of is insufficiently appreciated by most contemporary philosophy, it is a much more serious problem for him. Nagel, however oddly, can at least consistently maintain that selfishness is appropriate and legitimating when considering the structure of various political institutions, but Williams has the difficulty that to describe, if accurate, something as selfish is not only a critique, but a critique that he must regard as reasonable. He cannot in good conscience hold to some version of Nagel's distinction because of his position on the necessarily evaluative nature of normal discourse, yet, in denying that there are proper limits to political disputes, he seems to be saying that a particular evaluative category, that of the unreasonable, is, at least with respect to politics, inadmissible.
This is odd, because Williams recognises precisely this problem in the communitarian critique of post-Rawlsian liberal thought. Such a critique, as he does, urges attention to a particular community's linguistic and conceptual resources, claiming that in the absence of proper attention to them, a process of alienation occurs: a fully ethical life becomes impossible, as the evaluative categories necessary to sustain become attenuated and disappear. MacIntyre's infamous piece of absolutely hyperbolic rhetoric comparing the disappearance of various ethical traditions to an imagined destruction of the capacity to understand modern chemistry is perhaps the best example of this.
The problem Williams finds with such a critique is that it, in a way, denies the possibility of its own existence. For such a critique, he says in 'Pluralism, Community and Left Wittgensteinianism',
[t]he redirection of an ethical term, or more generally, the radical departure from an ethical practice, looks as though it will be merely arbitrary unless it can carry... a substantive body of agreement with it; indeed, some critics of the picture might say that even this is a kindly understatement, and that the consequence of the picture is that no change in practice comprehensibly occur unless it has already occurred - that is to say, it cannot occur at all except by magic.
The critique reifies practice to such an extent that it denies the obvious fact of ethical change, and because of that, if it is to position itself as critique of ethical change, the existence of what it purports to criticise. To put it another way, such a critique, because of the attention it demands for whatever ethical practices there are, can hardly critique exisiting ethical practices as deviant: they are precisely the ethical practices it ought to be demanding attention for. As Williams puts it:
[T]he Wittgensteinian is himself criticizing something, if only the practices of ethical theorists... part of our ethical practice consists precisely in this, that people have found in it resources with which to criticize their society. Practice is not just the practice of practice, but also the practice of criticism.
This, however, works against him too. Indisputably, people have coherently spoken of proper limits to political disagreement, primarily by attempting to articulate limits on the morally acceptable exercise of power. The area of legitimate dispute is not infinite, even within the bounds of the claim to authority rather than simple power, an area I fear Williams would find difficult to demark anyway for reasons essentially similar to those which cause him to struggle with the claim that no disagreement is necessarily illegitimate. For all that politics is about winning and losing, there are means to and forms of victory we rule out of court. Clausewitz's axiom cannot be, in its normal sense, regulative.