Either, providing it's properly designed. If the capacitive dropper supply uses low value Y1 capacitors for isolation, with proper clearances on the PCB, then it's no more dangerous than a transformer.
Wrong, capacitors do not provide isolation as already stated. This should be entirely obvious since the voltage at the output of a capacitive supply is still referenced to ground.
As for the touching the output of such a supply, you are clearly either very trusting or can't see the inherent danger. Just consider what happens if live and neutral are reversed with the typical resiator/capacitor/rectifier/zener design...suddenly the output is at full mains potential w.r.t ground.
Wrong, Y1 capacitors are connected across the primary and secondary in most switched mode power supplies. They provide more than adequate isolation at 50 to 60Hz.
The presence of Y1 caps across the primary and secondary creates a direct path from the AC main inlet to the DC output, ergo it's no longer truly isolated, albeit the maximum leakage current should be safe. However, since the leakage current is limited to a very small value, it's not actually going to be very useful for powering anything (micropower circuits maybe). Up the capacitor value to get more current and we are back to where we started.
Strawman detected! You've just leapt from "any capacitative solution" to "the standard capacitative dropper". Anyone reasonable agrees that a standard capacitative dropper does not provide isolation.
I think the OP's question is; what about the idea of a (hypothetical) topology that did provide galvanic isolation from primary to secondary, using only capacitors. Two interesting/reasonable (albeit very low power/inefficient/expensive) proposals have already been offered:
- Push-pull through pair of Y caps to be rectified on the secondary side (which is safety earthed on the secondary side to discharge the AC current that would come through those Y caps, just like the noise suppression caps in a traditional magnetic converter.)
Still not isolated, but now you've added a ground connection to guarantee the output is referenced to ground. That will get rid of those pesky leakage currents, but what would happen if you tried to e.g. float the output on top of another voltage which is ground referenced? Bang. If the output was isolated you could do this with impunity (obviously limited by the maximum isolation voltage in the supply).
- Relays/solid state switches that charge up a large cap on the primary side, and then "hand it over" to the secondary side. Fascinating concept; achieving safety ratings on those Relays/solid state switches would be "challenging" though.
Congratulations, you just (re)invented a flying capacitor converter
There are COTS off-line switcher ICs that do just this, but you are then at the mercy of a semiconductor device between you and 110/230VAC. Can you absolutely guarantee that device will fail in a safe state?
I'm honestly quite surprised at how many people don't understand what galvanic isolation is. As a simple thought experiment, take the output of your proposed supply and connect one or the other output terminals to ground referenced voltage (e.g. 240v live terminal). If it's galvanically isolated then nothing exciting will happen and no significant additional current will flow.