To answer the "why", floating point numbers offer a good compromise between speed and storage size vs accuracy. There are alternatives with different tradeoffs (e.g. Math::BigFloat, Perl 6).