Usually recruit a large study group, randomly give drug to some and placebo to others (control group), then follow them for an extended time to see if any statistically significant difference in rate/severity of disease or side effects between drug and placebo groups. Subjects don't know if they got drug or placebo (blind) and sometimes neither do the medical personnel who have contact w/ the subjects (double-blind).
Before doing a large scale blinded study like that, usually administer vaccine to a small group to test safety and see if vaccine is producing the desired antibodies etc. Unclear to me if the S African trials are a small-scale safety test or a full-scale efficacy test - kind of seems like the former.
Looks like none of the AIDS vaccine trials so far have been successful, one was halted because the drug group was actually suffering higher rates of AIDS than the placebo group.
Article discoussing why AIDS vaccine is so difficult
http://www.msnbc.msn.com/id/29898087/
Article discussing a possibly promising approach
http://www.sciencedaily.com/releases/2009/03/090312114801.htm
Appears that well over $1BN/yr is being spent on AIDS vaccine research.
Not clear to me why S Africa government cut off funding for the S African vaccine trials. Weird politics around AIDS in that country - the govt spent a decade denying it was a disease and/or claiming it could be treated with lemon juice - supposedly they've seen the light, but maybe not so.