C99's hexadecimal floating-point literals provide a convenient and precise way to unambiguously specify a particular floating-point value.  This capability would be a useful addition to Java; e.g. the documentation of the constants in Float and Double could use user-interpretible values rather than an opaque integer string.  One new method would be required in Float and Double, toHexString, and the parse{Float, Double} and valueOf methods would be have modified accordingly to recognize the new strings.