Bayesian confirmation by uncertain evidence: A reply to Huber (2005)