Arbitrary precision is not fit for databases because databases benefit enormously from having constant offsets for records and fields. As a practical matter, a

Arbitrary precision is not fit for databases because databases benefit enormously from having constant offsets for records and fields. As a practical matter, any columns in a database will tend to have values with a defined range of magnitudes and required precision. It's not like you're throwing arbitrary values into there; e.g., you're not going to be mixing high precision atomic measurements with interstellar parsecs and ounces of cocktail liquors. That doesn't mean databases must not support floats/arbitrary precision, but I wager most use cases would favor fixed point.