This concept is (at least in some respects) another step in a long conflict between those theorists who believe that threatened use of force cannot be extended usefully past
deterrence, and those who believe in another level called (for want of a better term)
compellence.
The fundamental difference lies not in the method of application of force, but in the aims of the actor appying it. First and foremost, that actor must ask: what is the goal? That would be my question to those proposing the above tactic. What is the goal?
It could be argued that the goal would be the use of nuclear arms to punish while minimizing the slaughter of the civilian population of the target. While this sounds well and good on the face of it, I am forced to ask: why? If the logic trail is worked back a bit, the answer is fairly quickly reduced to 'it is a more politically acceptable way to use nuclear weapons to do something that can be done without them, but which they make cheaper to do.' Even disregarding the dangers to stability and what strategic norms of restraint do exist, a DSNS as described here is simply a means of destroying a city without having to send lots of people and planes to do the same job, while (hopefully) giving the population time to evacuate.
Realistically, the panic that would occur if such a tactic were employed would likely kill enormous numbers of people anyway. There will be huge numbers who cannot or will not leave, no matter what you do, because (partly due to those norms) they will refuse to believe that the actor will actually 'do it' and detonate a weapon on a population...until it has actually been done, at least once, in recent memory. Indeed, the 'run up' to the DSNS could, it might be argued, serve instead to weaken the actor's case that they were willing to do so, demonstrating instead an attempt to 'bluff' the population out of the way.
Even if the population does leave the city, where will they go? Very few urban areas in the modern world have system infrastructure in their surrounding areas able to handle the load of the city center population without structural collapse, especially once the damage from the nuclear strike is done to the system center. At that point, rather than incinerated population, you have starving, diseased, suffering population. Which is worse?
The U.S. has demonstrated (several times, unfortunately) that it is indeed possible to collapse a civilian infrastructure (if not a military one) and, indeed, destroy whole cities via conventional bombing if enough effort and time is taken. The clear intent of a large-scale bombing campaign delivers a message of purpose that 'demonstrative' detonations cannot, despite their greater ferocity - it's harder to convince oneself that high explosive or incendiary weapons actually falling on one's neighborhood are a bluff. The time taken to do the thing right offers the same opportunities for evacuation that a nuclear strike does. The damage done can be customized to manage damage to infrastructure required to support a refugee population (water treatment plants, power plants, roadways for relief efforts, etc.) whereas a nuclear strike mostly likely cannot. Finally, the effort involved does impose costs on the actor - and I cannot say that that is a bad thing. If as a nation the U.S. (my country) is going to contemplate undertaking this sort of punishment of another population (and I'm not saying there are no conceivable reasons to do it!) I would be much more comfortable knowing that there was a high cost to doing so, and that the American polity had to decide to shoulder it in order to take the action.
The DSNS sounds, to me, suspiciously like the 'nuclear bunker buster' - a back-ported reason to continue to have and fund new nuclear weapons programs; a reason to cling to nuclear weapons as an instrument of compellent power rather than classic deterrent power. It is clear, after 9/11, that a nuclear arsenal is of little use (in the current strategic climate) in deterring non-state actors from acts of warfare against civilian targets. While the DSNS might, conceivably, be painted as a means of 'leveling the playing field' the costs of doing so are so high that any 'victory' so gained would be Pyrrhic at best. No; rather, it feels like an attempt to wrap some Congressional committee in a warm cloak of false nuclear coziness and safety blanket.
On a more concrete level, there are dangers to the actor inherent in this tactic. 100km high nuclear detonations do nasty, long-range EMP damage. Unless wide-scale disruption of this sort was planned, this would entail the risk of affecting third-party space assets and orbital/surface electronics, drawing additional actors into the conflict - during one such exoatmospheric test above Johnston Island, the U.S. managed to shut down electrical power to Hawaii and parts of California, and disrupt radio communications around the entire Pacific Rim for over eight hours. Multiple high-altitude detonations would involve low-volume but long-range fallout and wide-reaching EMP and scintillation effects.
Finally, there is a powerful norm against the use of nuclear weapons in combat that has been in effect since 1945. While it has been severely tested at times since then, it has so far held. Actual detonation of a nuclear weapon in combat may have (although this analyst is convinced it will have) a long-term 'genie from the bottle' effect, dramatically lowering the threshold for future nuclear use. For a nation worried about proliferators, that seems...counterproductive.