character kerning is bad anyway ...
I'm using it to get a paragraph one line shorter or longer, or to tune what in TeX is set with \parfillskip. I'm using ConTeXt with a mindset developed on other typesetting softwares. That's not right, i know, but for now i'm finding easier to work with character kerning than with \looseness, \parfillskip and \emergencystretch. \looseness=-1 quite never reduces the lines' number, because the inter-word spaces are already very well optimized. \looseness=1 quite always produces a short last line, so that i often work on \emergencystretch and \parfillskip to make it wider. But this way i have to work on 3 parameters; with kerning you have only one: uglier for ligatures but easier. I suspect TeX has better ways to deal with that problem, but i don't know them (hints are welcome).
effe
and this is
e ff e
you definitely don't want to kern like
e kern ff kern ff
but
e kern f kern f kern e
so, ligatures are pretty useless with intercharacter kerning (as per definecharacterning) ... just conceptually incomatible
(it is possible to turn it on but it really isnot what you want)
yes, the ligature would not get the same shrinkage or stretch of the rest, but i don't modify the kerning that much (factor=.02 is the maximum value). It's still acceptable. But i should consider all the text parts with a modified kerning and disable kerning around character sequences like "ff", "ffi", "fl" and so on. Luckily, the XHTML markup "knows" where the kerning is modified: <p class="wide">A paragraph with a modified kerning</p> should become <p class="wide">A paragraph with a modi<dk>fi</dk>ed kerning</p> Where <dk>...</dk> (dk=Disable Kerning) are tags to be inserted automatically before feeding the XHTML into ConTeXt. I'm not very happy or proud about it, but it can be done. Do you think it can be better done inside ConTeXt? Massimiliano