On Tuesday 23 October 2012 12:22:00 Mike Speciner wrote:
The character pair kerning you describe is what is used for text, and is typically NOT what you want to do for typesetting mathematics.
Sure. (If when you said "the last I looked ... [TeX] didn't have any [kerning]" you meant that it didn't have any *mathematics-specific* kerning, then I totally misunderstood; sorry!) [SNIP: brief description of CTI's optical kerning algorithm] OK, I understand. The devil, of course, is in the details; for instance, for either text or mathematics I'm pretty sure it wouldn't work well to kern everything so as to keep the minimum distance between one character's right boundary and the next's left boundary constant. As you say, "some smoothing and limiting", and probably plenty more... -- g