> So does Heirloom troff's implementation work by using newly > created tables akin to the ones for kerning? Or does it use > an algorithm to generate an intelligent guess?
You have to enter the values manually, using a "kernpair" request. (You decide which font you're pairing with which other font, and the combinations are practically limitless. I guess if cross-font kerning capabilities were more widespread, tables for download could be prepared for the most commonly used combinations, but ...)
