Normalization of rational functions

I thought this has been fixed some time ago:

   (19:36) gp > t; (-t0*t^2 + 2*t - t0)/(-t^2 - 1)
   %1 = (-t0*t^2 + 2*t - t0)/(-t^2 - 1)

Should not the normalization choose the positive sign for the
leading(?)  coefficient of the denominator?  (Or at least the
numerator?)

Thanks,
Ilya