I've been in contact with Mike Cowlishaw of IBM.
He has rewriten the java class BigDecimal (used in the DecimalFormat).
He implemented the new class by using decimal string math instead of relying
on the processor (hardware) specific math.
http://www2.hursley.ibm.com/decimalj/.
The BigDecimal class decimal and floating point aritmetics with unlimited
precisio.
This new implementation will probably become part of the Java math package.
He is also writing the same class in C.
The class should be easily portable for the fact that does string
math.
Now the ICU has no BigDecimal class in C++. The math is done in double and
the BigDecimal code is commented. The double has a limited precision and use
floating point arithmetic that is not suited for business applications.
I think that the new class, developped by Mike Cowlishaw should be possibly
included in the ICU.
