smryan@garth.UUCP (Steven Ryan) (01/08/91)
Followups to: comp.lang.misc > . . . The famous "NaN"---which is > *allowed* to be *undetected* by ANSI FORTRAN 77 Standard. Am I correct? No. Fortran constrains but does not define the real arithmetic. Questions like these are simply not addressed by Fortran. > If you know why, please enlighten me. Is this kind of thing specified in > FORTRAN 9x? > >Alas no. Language committees still mostly punt on FP issues. There is As well they should. There are many issues which are not part of a language definition, and instead should be factorred into different standards. IEEE arithmetic is merely one possible model for real arithmetic. Some machines implement very different models which are still valid for Fortran. Perhaps someday there will be a model that blows IEEE away and is still valid for Fortran. By separating the algebra from the model, Fortran maintains a degree of machine independence. (Which was the purpose of standards when I was younger--a standard algebra, not a standard model.) The same factorring occurs in integer arithmetic, character set, IO interfaces, libraries, memory allocation, ... -- ...!uunet!ingr!apd!smryan Steven Ryan ...!{apple|pyramid}!garth!smryan 2400 Geng Road, Palo Alto, CA