Raymond writes:
> Unfortunate in some ways. At least with the macros, you had some idea
> that the author thought the numbers should be "small". Now you
> don't.
True, but you have to recheck that anyway to be safe.
> However, this is particularly bad with gcl (and perhaps
> others) that just believe you and truncate the numbers to get wrong
> results.
It often depends on the optimization level, I suppose.
> Andreas> Asking the user if he really wants to do a calculation if
> Andreas> certain limits are reached might nevertheless be a good
> Andreas> option as Richard suggested.
>
> Yes, as long as the limits can be set by the user.
Yes, of course. This *must* be user settable.
> How is this idea supposed to work with the trend of having maxima not
> asking any questions at all anymore? Just give an error and exit?
I'm not so convinced that this trend is good :-(
I prefer the software asking me questions instead of silently proceeding
to compute the wrong answers.
'Andreas