There's some interesting discussion over on Jonathan Ichikawa's blog about the doing/allowing distinction. Jonathan denies that this distinction has any great moral significance. (Call this the "Moral Equivalence Thesis", or MET.) In defending this position, he makes an important point: those who would attack MET need to first make sure that their examples actually involve corresponding (in)actions.
For example, it's no use pointing out that murdering someone is much worse than failing to prevent them from walking into the street where the murderer secretly lay in wait. For here the 'doing' and the 'allowing' do not match up. The action is an intentional killing, whereas the inaction is merely aimed at a walk in the street. As Jonathan explains: "The apt comparison is between (knowingly) killing someone and (knowingly) refraining from preventing her death."
I've always thought the greatest worry for MET was its disturbing implication that we're all responsible for the unnecessary deaths in third world countries (whenever we fail to give to Oxfam). But what would the 'action' corresponding to this 'inaction' be? The inaction seems ubiquitous in our lives, rarely considered, and largely necessary for our luxurious standard of living (though of course occasional sacrifices would not be too trying for us). Moreover, it is only very indirectly related to the eventual deaths.
Bearing these facts in mind, I think the corresponding 'action' would go something like this: Imagine our 'day-to-day' bank account is set up to transfer its contents to Oxfam at the end of each month. (Perhaps at the start of the month it gets refilled from our savings account, or some of our income is deposited straight into it, or whatever.) Then, whenever you spend money from this account, you are acting in a way that reduces the amount of money Oxfam receives from you. (You know this well enough, though you might rarely give it a second thought.) Your actions indirectly cause more people in third world countries to die; just like your actual inaction does. It seems clear that the two are morally equivalent.
I think MET remains extremely plausible once clarified in the above scenario. When we take care to ensure that the actions and inactions (or 'doings' and 'allowings') are genuinely corresponding, then we find that MET is a much less radical thesis than we might originally have thought. The action described above is not morally equivalent to directly killing people, at least not without adding many further assumptions. So MET alone does not imply that we might as well have shot a poor person when we failed to give to charity.
No, even in this extreme case, when MET is properly understood, it appears to be quite plainly true. And this is even more obvious, I think, in cases like euthanasia. Denying MET here can lead to bizarre results. We might end up with life-support machines that are built to automatically turn themselves off unless the operators press a button. That way, the patient's death would be purely due to the doctor's "inaction", and not his "actions". But I'm sure you'll all agree that this is a superficial difference, and that it couldn't be any better or worse to flick the switch off than to refrain from pressing the "stay on" button.
And this superficiality is precisely what MET recognizes. Whether technically "action" or "inaction", "doing" or "allowing", it's really all the same.
(A speculative final note: I think this ties in with the "baseline" issue discussed in my post on framing thought experiments. Perhaps "actions" are seen as deviating from the 'natural way of things', whereas "inaction" simply leaves things at the baseline. If so, the lack of any metaphysical grounding for this ghostly notion of a 'true baseline' should provide us with further grounds for doubting the significance of the doing/allowing distinction.)