We show that a discrete infinite distribution with finite entropy cann
ot be estimated consistently in information divergence. As a corollary
we get that there is no universal source code for an infinite source
alphabet over the class of all discrete memoryless sources with finite
entropy.