No subject

[M. Douglas McIlroy]

Contrary to rumor, bankers often use floating point.

1. For higher financial computations, of which continously
compounded interest is a simple example, and portfolio
management is a complicated one.

2. To represent numbers that need more bits than most machine
integers hold.  Roundoff is no problem if you count pennies.

Cobol has no monoopoly on financial programs.  Various
outfits use Fortran, C, and other languages that do 
not support big enough integers, either decimal or binary.

Even Cobol has "computational" data types that are customarily
floating point.  Depending on the frequency of 
base conversion relative to other operations, it can 
be more efficient than decimal-string arithmetic.
And, just as in scientific computing, it alleviates
concerns about scaling.