Seeking newby advice for setting up kerning classes

Firstly, note that that the concept of key glyphs in classes—what I think you mean by ‘master characters’—is specific to FontLab Studio 5, and does not occur in other tools, including newer versions of FontLab. The key glyph concept was actually pretty useful if one knew how to work with it, and I occasionally wish it had been maintained in FontLab, because converting kerning from a single glyph to a class—e.g. when one adds a new diacritic letter to a font—is frustrating as currently implemented. But key glyphs are not part of the OpenType GPOS pairpos class format, so the way FontLab 8 and other tools work is closer to how the output fonts work.

My approach is to define 1st and 2nd kern classes* based on general shape of the left and right sides of glyphs within a primary height zone. This latter concept isn’t really discussed much, but I think it is critical to keep in mind throughout the class assignments and then the kerning. So, for example, I consider the primary height zone for uppercase letters to be between the baseline and the cap height. So in many designs, D O Q and related diacritics go together in a 1st class, and C G O Q go together in a 2nd class, because these tend to share similar shape and sidebearings, respectively right and left, within the primary height zone.

Obviously, in that example, the tail of the Q is outside the primary height zone, and that is what you keep in mind when you are kerning, because it will produce a number of exceptions to the class kerning.

Narrow letters with diacritics above require a decision about whether to include in classes. I tend not to, because the number of exceptions when they follow T V W Y f etc. and relative to quotation marks, parentheses, or other punctuation tends to exceed the number of useful class kerns. So typically I do not put ì í î ï ǐ ĩ etc in classes, but kern them individually. [Recently, I have started putting /i.dotless into a class with î on the basis that this dotless variant is only going to be output from GSUB when it is followed by an above combining mark, so I want a method to kern it away from T V W Y f etc..]

In terms of making the classes, once you have done it for one font, those classes can easily be copied to other projects with similar character sets, and then trimmed or extended as needed. I do encourage you to try to establish your classes as fully as possible before you start kerning, because converting kerning data to or from class kerning can be tricky. An important thing to bear in mind is that if you discover during the kerning that you need to split a class, e.g. the shape of the right side of D is not similar enough to O to be in the same class and is requiring too many exceptions, you need to first flatten the class kerning to pairs, then create any new class and move glyphs from one class to another, and then compress to class kerning to get existing kerning properly assigned to the revised classes. [I am describing a process to use in FontLab. What Glyphs does when compressing to class kerning appears to erase all exception kerns, which is terrifying and not useful.]

There are a lot of well-made open source fonts now available, so you could download the source for one of those, open it in your font tool, and examine the class assignments. This can be more useful in understanding that trying to read a manual.

_____

*Note that I have spoken here about 1st and 2nd classes, rather than left or right classes. That’s my preferred way of thinking, because I am sometimes making fonts for bidirectional writing systems, and left is not always 1st and right is not always 2nd. Tools like FontLab and Glyphs use left and right designations, and if you’re only making Latin fonts it is fine to think in those terms.