Floating-point related issues

Throughout the Flockwave protocol, we mostly avoid using floating-point numbers, except in places where we expect the exponent of a value in standard scientific notation to cover a wide range (several order of magnitudes). This is intentional, due to the problems that one may encounter with floating-point numbers in almost all computer languages from low-level ones like C to high-level ones like JavaScript.

As an illustration, consider what happens in Python, a popular high-level language when attempting to serialize the value 0.3 (expressed as 0.1 + 0.2) into JSON:

The net consequence of this is that any time Python would attempt to serialize a floating-point number into JSON, we cannot be sure that it does not end up with an unnecessary amount of extra digits at the end due to rounding errors. Therefore, in cases where it is possible, it makes more sense to use signed or unsigned integers when transmitting these values over the wire, and then do the appropriate conversion to floats at the very end of the data processing pipeline where the value has to be presented to the user.

For more information about this topic, refer to What Every Programmer Should Know About Floating-Point Arithmetic for a high-level overview and What Every Computer Scientist Should Know About Floating-Point Arithmetic for a more thorough academic discussion.