Interesting - from the first link, we see that any given fold can be encoded by some 10^10 different serquences.
"If we exclude detectable homology -for a sequence of length 100, on the order of 10^113 sequences exist that are > 20 % identical - we still remain with 10^10 distinct, foldable sequences
for each fold."
Emphasis mine.
That site also mentions the bit about random sequences (TB was correct). Of course, when one neglects to mention the parameters, big numbers are impressive. When one considers that selection is not random, and that genes encoding useful proteins would be selected for, those numbers are moot.