I think I have to rule the 40 character solution out of bounds on the basis of the fact that you are walking through the numbers and testing whether each is prime. This loses the central idea of Eratosthenes which is that when you find a prime you immediately mark off its multiples.

Basically you reversed the role of the 2 loops.

The 60 character answer is impressive. However it seems to scale quadratically. I must confess that I don't see why it is scaling quadratically, but it clearly is. So unless you can explain why this is a bug in Perl, I am going to have to call this a solution to only the first problem.

UPDATE
I now understand why your 60 character solution is scaling slowly. It actually scales like O(n*n/log(n)), and it is because you are walking all of the primes with each strike rather than just the multiples of the current prime. So it definitely only is a solution to the first problem.