Noone can do this exactly, but why isn't some approximation effective? To update in a Bayesian way we need to know our priors too, and not being able to state the numbers precisely isn't seen as a reason for not using Bayesian updating in a wide class of practical situations.