C language compilers data type

There are at least two types of integers (short or long) and two types of decimals (float or double) in most C language compilers. We should write all the programs using longs and doubles without ever needing shorts or floats. What do you think was the idea of adding these two additional (smaller) types?