We study a problem of quantization of a discrete probability distribution 
for an application that transmits it once. Consequently, instead of minimizing 
the expected distance to reconstruction points, we minimize the worst case distance. 
We show, that in such setting, the quantization problem turns into a variant of 
the covering radius problem, where the space to be covered is a probability simplex. 
We use this connection for derivation of asymptotic (high-rate regime) expression 
for achievable covering radius. We also propose a simple practical algorithm for 
performing quantization of discrete probability distributions. We show that this 
algorithm is optimal in the asymptotic sense.