Python float doesnt have enough precision to distinguish a difference at around 1e-16, so in Python, 99...9 (as you wrote) is stored equal to 10...00 (with as many 0s as there were 9s above).

Interactive shell:

Input: float(10000000000000000)==float(9999999999999999)
Output: True

You can also use the decimal module in python for accurate floating point arithmetic :

from decimal import Decimal
print(Decimal(9999999999999999)/Decimal(2))