Probably the first step would be a good, non-regular-expression (IIRC, URLs are not regular) based link parser that's ported to every possible language on the planet.