Hi
In my opinion, one of basic solution is that in addition to any type of
floating point real type, system will provide storage for string type. The
possible anytype(include integer-fraction) number is put in the storage.
For example
(8/9): Real type[4 or 8 byte] + string type[8][/][9]
(1.0): Real type[4 or 8 byte] + string type[1][0][/][1][0]
(3.14159): Real type[4 or 8 byte] + string
type[3][1][4][1][5][9][/][1][0][0][0][0][0]
If the system need to conversion float to bfloat, system can refer string
type for avoiding information-drop. But this method has few problems.
1)The system needs more big memory.
2)On the way to calculating, how treat irrational.
3)Xmaxima develop-team needs the large scale remodeling and system checking.
Ken