Questionable classes, more control + written documentation ?

Post Reply
Posts: 2
Joined: 01 Mar 2023

Questionable classes, more control + written documentation ?

Post by Gagaramond »

Hi Tim,

I'm wondering about some of the automatically generated classes Kern On produces.
For example, my f_f has an f component on the right, yet KO puts the right side in a KO_f_f class while f is in the KO_f class. In this case the auto pair for f_f is the same as the model though.

Another one that bothers me even more, though I understand it because of the different shapes : period and comma don't end up in the same kerning classes, leading to different kerning which I absolutely want to avoid (I don't want text to reflow differently between period and comma, except maybe for some exceptions like g,

Is there a specific reason/logic I'm missing for that, wouldn't it save Kb to put them in the same class?

I would like a technique to avoid this and force the algorithm to associate glyphs in the same class. I've tried putting more same model values but it doesn't seem to change the resulting classes. I've also toyed with the kb value over the advised 64 (?) limit, but whatever value, I always get a font of 94Ko with the same classes for these examples.

It would be great if there was a way to "fix" that, or setup some fixed classes in advance. I think you'll agree that correcting classes manually afterwards is not a workable solution since KO regenerates classes every time.

Additionally, it would be great to have a written documentation or FAQ to refer to. The videos are great as an intro, but with my bad aging memory, I'm regularly trying to find a specific information with the search function. I had to look into the forum or twitter to figure things out. Maybe some things could be compiled in a clear form on a page ?
For example explaining the Kb limit somewhere with things to take into consideration when choosing a value ?

Thank you
Post Reply