6.—“No state at war with another shall countenance such modes of hostility as would make mutual confidence impossible in a subsequent state of peace: such are the employment of assassins (percussores) or of poisoners (venefici), breaches of capitulation, the instigating and making use of treachery (perduellio) in the hostile state.”
These are dishonourable stratagems. For some kind of confidence in the disposition of the enemy must exist even in the midst of war, as otherwise peace could not be concluded, and the hostilities would pass into a war of extermination (bellum internecinum). War, however, is only our wretched expedient of asserting a right by force, an expedient adopted in the state of nature, where no court of justice exists which could settle the matter in dispute. In circumstances like these, neither of the two parties can be called an unjust enemy, because this form of speech presupposes a legal decision: the issue of the conflict—just as in the case of the so-called judgments of God—decides on which side right is. Between states, however, no punitive war (bellum punitivum) is thinkable, because between them a relation of superior and inferior does not exist. Whence it follows that a war of extermination, where the process of annihilation would strike both parties at once and all right as well, would bring about perpetual peace only in the great graveyard of the human race. Such a war then, and therefore also the use of all means which lead to it, must be absolutely forbidden. That the methods just mentioned do inevitably lead to this result is obvious from the fact that these infernal arts, already vile in themselves, on coming into use, are not long confined to the sphere of war. Take, for example, the use of spies (uti exploratoribus). Here only the dishonesty of others is made use of; but vices such as these, when once encouraged, cannot in the nature of things be stamped out and would be carried over into the state of peace, where their presence would be utterly destructive to the purpose of that state.--Kant Toward Perpetual Peace: a Philosophical Blueprint, translated by Mary Campbell Smith. [8:346-7)]
The above is one of my favorite passages in a work, Toward Perpetual Peace, that I am very fond of (recall this post, where I situate Kant in a tradition that includes Spinoza and Hume, but also public choice theory). The underlying strategy -- and this accounts, I think, for the texts astounding influence on real politics -- is to argue for very idealistic ideals from extremely thin or even pessimistic (Augustinian-Hobbesian) commitments about human nature. At no point does Kant rely on a reform of human nature. But that's not the only strategy he deploys, and here I want to call attention to another.
Remember that for Hobbes, the whole point of generating a sovereign is to escape the state of nature by which Hobbes means both the state of war and the permanent preparation of war. Unlike Master Mo (recall), Hobbes does not really propose a solution to the problem of the existence of the state of nature at the level of international relations. Kant is aware of the problem because he offers an argument against the existence of standing armies (a much debated issue in the eighteenth century -- Kant sides with the republican preference for militias) because these basically signal that one has not escaped the state of nature.
Okay, let me turn to the quoted passage. You may think that Kant is offering an argument from enlightened self-interest to the conclusion that one must be willing to restrict voluntarily one's conduct in war. This is not all wrong, for one with such voluntary restrictions on one's conduct one signals to all belligerent parties (including one's own side) that one is waging war in the service of peace and so can be treated as a serious negotiating/peace partner when the time is ripe.
But there is a more important, perhaps more subtle point, lurking in Kant's argument. The worst version of the state of nature just is total war or a war of (mutual) extermination. This is what we wish to avoid when we contract in, or recognize the legitimacy of a sovereign. So, to engage in total war is to undermine the very reason for the existence of sovereign states. And this remains true even during the conduct of interstate war. A war of extermination is to be avoided at all cost because it undermines the very rationale for the existence and legitimacy of a state (and it invites excessive retaliation). So, if you want to avoid at least one of the known causes of a war of extermination, which just is the worst version of the state of nature, you need to show restraint in war. (And this, in turn, signals that you care about escaping the state of nature, and wish for peace.)
The underlying idea here is to take Hobbes's idea about the escape from the state of nature seriously and to adhere to that even in the conduct of war (and so to prepare a durable escape from the state of nature). I think that's pretty convincing.
But you may think that if you are not on the receiving end of a war of extermination (but say on the winning side), you may not really have returned to the state of nature (you are being protected by your soldiers). Let's grant -- for the sake of argument -- that this is true sometimes, but it is rare one can fully escape the effects of war altogether (although sometimes these effects are indirect). And the whole point of sovereignty is to avoid the state of war.
Kant himself has another response to the objection. He thinks the internal constitution of a state matters quite a bit. He calls it a republican state, but it's clear this is a republican theory influenced by Montesquieu (or Madison [it's clear that in the text he treats the establishment of the United States as a possibility proof of continental size republics], or Hume) and less Rousseau. For, Kant insists that his blueprint requires states with separation of powers, a full franchise, representative government, and commercial relations with the outside world; he is very clear that Rousseau's version of republicanism leads to tyranny--not a surprising claim in 1795. (This is really the outlines of liberal democracy.) In such republics the costs of war, even a winning war of extermination are felt by those people who need to consult. This is, in fact, why even though Kant rejects standing armies in the best circumstances (the confederations of peaceful liberal democracies), conscripted armies are better (than, say, our own professional armies) during war. The cost of war is distributed widely.* There is an empirical-psychological question to what degree this will really reduce the willingness to consent to war. But here the point is a normative one: in such circumstances one cannot escape being pushed back to some degree into the state of nature even if one is on the winning side of a war of extermination.
*He also rejects debt-financing of war because then the cost of war is shifted on to creditors and future generations.