From Keith Arnaud: The reactions I recall to my comments about saving posterior distributions were that Jeff Scargle was very sceptical about using other peoples posterior distributions as priors since it would lead to errors propagating. He preferred neutral priors. Then Tom Loredo noted that it was also important to know the likelihood functions used. If you took posteriors on two parameters from different experiments and combined them as a prior and the two experiments themselves used the same prior then you would include this prior twice.