Indeed it is possible. (Even if pi is not normal to any base.) Now I'm curious how *probable* it is, or complementarily, how probable it is that at least one finite string is missing from an infinite string of digits chosen at random. When doing calculations like this I tend to get stuck when trying to account for substrings that overlap. At least for finite calculations. Hmm, the chance of missing a given single digit d is 1 - (1/10)^oo. The chance of missing a single pair of consecutive digits (d, e) is ??? Maybe it's best to *estimate* the chance of N consecutive digits chosen independently at random having every string of length k at least once (for N = N(k) >> 0), and then let N —> oo ??? —Dan David Wilson wrote ----- Presuming pi is normal in all bases, aren't all possible (finite) messages somewhere in there? -----