yeah, main things are tracking overlay collision, groups like radio buttons, where you start your measurements from top left or bottom, dpi, and zoom levels. built systems for it using itext and websupergoo abcpdf. (backend for manipulating pdfs, tamper sealing etc.) and front end uses a image of the rendered pdf page that was extracted. This is a bit different from pdfjs but also scales well. PS pdf standards suck. :(