Chapter two takes on this question of distributive justice. SRM carries substantial risks of injustice, for it threatens disproportionate harm to persons who bear little or no responsibility for anthropogenic climate change, notably the global poor and future generations. I argue that SRM is best thought of in terms of so-called “non-ideal theory.” As opposed to ideal theory, which investigates what justice requires assuming full compliance with our duties of justice, non-ideal theory concerns what justice requires in cases of only partial compliance with such duties. It is plausible to view humanity’s failure to cut greenhouse gas emissions as a case of non-compliance with our duties of (climate) justice, and so it is likewise plausible to take a non-ideal approach to climate justice in general, which involves searching for policy options that are both politically feasible and likely to be effective in managing climate injustice. This gives SRM much more traction than it would otherwise have. In an ideal world, presumably we would have mitigated our emissions long ago, such that climate engineering would not be on the table. But in the actual, non-ideal world, we have at best only partially complied with our duty to mitigate our emissions, and the prospects for full compliance in the near future are dim. Accordingly, we need to consider what justice requires given the fact of partial compliance, and this is the province of non-ideal theory. I argue that, under certain conditions that could realistically hold in the future, some uses of SRM could be favored by non-ideal theory. This is because SRM has the potential to manage various injustices associated with climate change, such as by slowing the rate of warming and thus allowing time for emissions mitigation and adaptation to changing climatic conditions. This chapter also addresses compensation to parties harmed by SRM, which could go some way in alleviating unjust distributions. However, SRM compensation faces some difficulties, including technical challenges in detecting impacts and causally attributing them to SRM. Nonetheless, I make a case that a non-ideally just SRM policy must include a system to compensate victims of injustice, even if that system is imperfect.