I drink all my samples. The chances of contamination from the sample's contact with the test flask that you float the hydrometer is really really high unless you're diligent about sanitizing it. Also, particularly with wine, you can get pretty good at evaluating the progress of a fermentation by tasting the sample. Also - I'm usually just really curious about what the stuff tastes like.
As far as using a refractometer to measure SG, you just have to convert from "degrees balling" or Brix to SG. The easiest way is to look at a triple scale hydrometer. Its calibrated with SG and Brix and you can just read across the scales for a quick conversion. Also, you can Google hydrometers and get to these kinds of conversion tables.
http://www.ertco.com/hydrometer_conversion_tables.html
Now there's one other aspect that I forgot to mention. Hydrometers measure the density of a liquid in relation to the density of undiluted water. Refractometers are generally calibrated to measure relative concentrations of sucrose in a sample by measuring the deflection of light through the sample. This causes problems when trying to directly convert because 1.) the density of wort is effected by things other than sugar, and therefore hydrometer readings are always reflecting an artificially elevated sugar/particle content and 2.) wort contains maltose, not sucrose.
This first problem is really not that big a deal, and is apparently termed "Apparent extract v. Real extract", and is really a correction factor of only about 2-6%. Likewise the maltose conversion is a tiny number.
The upshot is that if you're looking for real precise SG readings, you're going to need to take both apparent-extract-corrected SG and Brix measurements and factor them together to get a true picture of the actual sugar content of the wort.
In the meantime, I'm not really that interested in getting that precise and although I've been taught how to do all that crap, I really am happy with a quick-and-dirty-and-get-back-to-my-RDWHAHB-pint once I figure I'm floating around the right ballpark. I tend to think the refractometer is a good tool because the sample is small and its a quick read for on-the-fly measurements and when to kill the boil, etc. but I generally rely on SG readings for comparing apples to apples when I calculate the alcohol content.
Sorry... long answer...