Closed forms of the optimal Bayesian design for the estimation of the log odds-ratio from a two-sample experiment are obtained. The effect of various specifications of prior distributions are considered and plausible circumstances where unequal allocation is optimal are presented.