First of all, thanks to Chris Connor for posting v.1 on the Make: Blog!
Now, some of the considerations I've come up with for v.2:
The lack of a good ferrite core may prove the undoing of this design. As it is, the ones that I have access to seem a bit large (over 1" long and .5" wide), and the more ferrite, the higher the flux density needed to achieve saturation, and therefore the higher the current needed to drive the dc-dc upconverter. This is also why I don't want to use a regular transformer. However, a small enough isolation transformer, like those used in VCRs, TVs, stereos, etc., might just do the trick.
Next time I'm building on a breadboard. Period.
The capacitor has to be limited to below 300 µF, since the upconverter runs at about 50 kHz. If you're wondering why this is, check out the concept of reactance.
Similarly, the original Joule Thief implementation of this upconverter ran a white LED. This means that they were using a half-wave rectifier on the circuit. If I manage to pump the voltage up to about 5.6 Vrms out (with a +2.8 V offset, meaning that the AC signal is all above the zero line), I'll be able to put a regular low-wattage diode in to trim some of the oscillations, and increase the size of the next capacitor downstream. However, I could make this an LED if I can get the design up to 6.7 Vrms.
Do I really need an LED in there, though? Somebody weigh in on this in the comments. I know that the higher the voltage (short of probably 8-9 Vrms) the better, so I'll build for this anyway, but it does seem to me that the LED might just be a waste of wattage, as might the rectifier itself.
Anyway, enough about v.2. I'd better get building.