Sergey Gavrilets and William R. Rice have a new population genetic model for homosexuality out. You can read the full paper over at Gavrilets' website, while Ars Technica and Matt both have some nice commentary. I don't have much to add, and generally share Matt's skepticism of the utility of a one locus model, but at least it is somewhere to start. I will reiterate that the "problem" is obligate homosexuality. I don't find facultative homosexuality as surprising or evolutionarily mysterious. Also, I am intrigued by the the older brother effect because of its implications of genomic conflict.