Arthur Reutenauer wrote:
when i convert the to TT format i get a correct output
If Luigi is concerned with rounding errors, he shouldn't do that: while the coordinates of the points on the curve wouldn't change, all the control points would need to be recalculated due to the conversion from cubic to quadratic splines. On the other hand, if you keep CFF outlines and reset the units per em size from 2048 to 1000, you're of course exposed to rounding errors, but all the original points are retained, both points on the curve and control points -- thus it's much less error-prone, in my opinion.
Rounding errors are not that important in this case, and there are downsides to both (o3->o2 touches *only* control points, so in some sense it is more precise than upm change). What's normally more important is that if you convert to ttf, you will loose all hinting information. Best wishes, Taco