Too loose compared to the exact same interaction elsewhere?

[brianbrubaker]

Really loving Kern On so far—thank you for your great work, Tim!

Do we know what's causing something like this to happen, where KO is suggesting that the interaction is too loose / too tight when it sometimes is literally the same exact interaction in a glyph with the same angle and side bearings?

In this instance, the issue seems to disappear after restarting Glyphs, but this happens frequently enough to mention it here.

[Eben Sorkin]

I am curious about this too.

[brianbrubaker]

Another example

[brianbrubaker]

Another example of this where both the Y and the A have the exact same side bearings and equal spacing. The KO model I have set from the A—Y is -15. However, the Y—A is auto kerned to -2. Do I need better models or does anyone have any idea what is causing these kind of inconsistencies?

[Eben Sorkin]

I suspect the answer, for now, may be that your models are inconsistent. But it would be great to be able to ask Kern-on about the source of the inconsistency so that it is faster to resolve.

[SCarewe]

The A–Y "model" you have set in your screenshot is not a model, it's set to Independent.

[brianbrubaker]

SCarewe wrote:
> The A–Y "model" you have set in your screenshot is not a model,
> it's set to Independent.

Indeed! A poorly timed screenshot, I admit. I set it to independent to start adjusting the kerning how I wanted it right before the screenshot. It was a model right before I took this, I promise ;-)

UPDATE: The problem with that interaction was solved by removing some other models. Both sides then resolved to the same spacing.

[brianbrubaker]

Eben Sorkin wrote:
> I suspect the answer, for now, may be that your models are inconsistent.

Yes, I've been able to solve a few things by going back and removing some models, and tweaking some others as in the case of the "AYA" example I posted. But the main issue here has still persisted.

> But it would be great to be able to ask Kern-on about the source of the
> inconsistency so that it is faster to resolve.

Agreed! It does often have suggestions about "this interaction seems to be tighter than this other interaction", but it doesn't help if those interactions are literally the same just reversed like in those screenshots I've shared.