Comment 2 for bug 548619

i see it would be an advantage when the files are processed with other software, e.g. a vector editing program.
however this is not easy to change.

arrows: possible, but protocol between parser and the rendering backends needs to be changed.

rectangles: this is would require something like a "optimizer", a processing after the images has been parsed. it would need to implement some algorithms to detect common shapes. rectangles are one example, fills would be an other (they are now made of many short lines even when they could be replaces by one long line).