We need support for UTF8 and shiftjis

There's Encode::Guess for when you need to determine automatically which of several encodings a file has. It sometimes even guesses correctly, in particular if you need to distinguish between only two encodings whose representations aren't too similar. But don't expect miracles.

Other than that, specifying the appropriate encoding layer for open() should work just fine. Have you tried shiftjis (instead of utf8) for the file where utf8 failed?