Actually, I think just adding your comments
> realroots does not look for (or find) approximations to complex roots. It
> only finds rational bounding intervals around each of the real roots, of
> width rootsepsilon, and then returns the center-points of those
> intervals.
to the documentation of realroots would be sufficient.
Who would be responsible for that?
-sen
On Fri, 25 Aug 2006, Richard Fateman wrote:
> Definitely, reading the documentation is advisable. Maybe the name of the
> program should be improved, though the classical name for the computation is
> "real root isolation".
>
> rootsepsilon influences the bound on permissible error.
>
> It uses Sturm sequences, or at least originally did so, to find isolating
> intervals. The refinement can be done in any number of ways, but I suspect
> something not very clever, like bisection, is used.
>
> Realroots will always compute rational numbers. In the commercial macsyma,
> if rootsepsilon is a floating-point number, the rational number is converted
> to float. similarly for bfloats.
>
> Perhaps the documentation should be made more explicit, leaving no doubt as
> to what is being computing, how and why.
>
> Raising another point: when the documentation is changed, do the versions
> in Portuguese, Spanish, German... also get fixed?
>
[snip]