The fact that arrays can grow to fit your needs is just one of the really great things with dynamic languages in general and with Perl in particular: no need to keep track of the size and to malloc new memory chunks.

Now, having said that, if you want to implement such a limitation, it is quite easy to do it: just don't access directly to your arrays, but do it through a small set of functions that check for bounds and return an exception when your program gets off-limits. So you could implement special versions of push, unshift to check bounds, as well as a set_val and a get_val mutator and accessor functions to access to the array elements.

Although I am not a great fan of OO programming in general, this is one case where it really makes sense and I might very probably implement it as an object-oriented module to create and manage arrays through a bunch of methods. (Or, perhaps, I would be tempted to do it with some closures, this usually makes it possible to do the thing in probably twice less code compared to OO code, but that's a different question and a matter of purely personal inclination.)

As mentioned by other monks before, a tied array might be a more transparent solution, but I haven't been really convinced so far by tied arrays and hashes.