Perhaps this might sound like a foolish suggestion, but...
Wouldn't it be better (for some definition of better) to design our
BigInt type so that we can replace the math engine with a different math
engine? Eg, so we could plug in GMP, or PARI, or Miracl, etc.?
(With a fallback to our own math
Based on my experence on implementing the original Big* packages
for perl, I have some suggestions after reading pdd 14.
I suggest that instead of storing the mantissa of a Bignum as
a BCD string of nibbles that it be stored as a string of 16 bit
values each holding 0... Note that this gives y
