Kerning class definitions

Titus Nemeth

My understanding is that it is considered best practice for OT class-based kerning to define kerning classes for everything (apart from exceptions) that one wants to kern. That includes classes which contain only a single glyph, and therefore don't require a class as such.

Can anyone confirm if this is still considered best practice to avoid kern subtable breaks? Or is it acceptable to kern single glyphs against classes?

Thanks!

Bahman Eslami

If your primary concern is subtable breaks, the useExtension keyword in FDK doesn't solve it?

Titus Nemeth

I don't have an actual implementation problem, the question is of a more fundamental concern: is it better (safer) to define kerning classes as I've outlined above (before running into problems), or has this become irrelevant (because of better compilation routines or whatever else).

Bahman Eslami

The only case I remember putting one glyph in classes made sense was to tag them with script (e.g. HE, AR) names for kernFeatuerWriter script in FDK. Tagging classes helps to separate the scripts in different lookups, which in turn also reduces the possibility of subtable breaks.

Georg Seifert

My attempts to write kern features produced better results when all glyphs have classes. 

And a different advantage is that you can easily spot exceptions. 

Titus Nemeth

Thanks for your reply Georg, this tallies with my information from a few years ago. It was specific to the AFDKO, so makes sense that it applies to Glyphs as well.