Axis Location Specifier: Understanding the Role of Plus and Minus Signs

[ErwinDenissen]

I’m curious about the necessity of this. While I can see how it might be useful in relation to axis range conditions (ConditionSets), that's something more relevant to Stage 2. Could you explain the reasoning behind it?

[skef]

Consider the archetypal discontinuous kerning problem: A designer is making a variable font with a weight axis. She is kerning T and o. At lighter weights she wants the o nestled under the top stroke of the T. At some weight the o no longer fits next to the vertical stem of the T and has to move out. And she does not want it to move gradually, as the intermediate positions look wrong to her.

She could simply use two different axis locations that are "close" to one another to simulate discontinuity, but that relies on no users picking the intermediate locations. The plus and minus signs make it easy to specify locations that are truly adjacent according to the specification.

With multiple axes things can get a little tricky, and choosing two positions that are "diagonally adjacent" in designspace may not be absolutely foolproof, but they will be as close to foolproof (e.g. not revealing any intermediate interpolations) as practical with the current spec.

[ErwinDenissen]

I see what you're getting at, and I understand the need for precision in avoiding unwanted intermediate positions. However, I believe feature variations for GPOS already address this scenario effectively.

[skef]

There are a number of scenarios that can be addressed either by adding "masters" (in the case of kerning, a value) into the designspace of a variable element or by substituting one (variable) element for another. Some designers choose to deal with problems like the number of bars in a dollar sign with additional masters.

A given foundry might choose to be prescriptive about which way to solve a problem, but generally speaking we want to support the different patterns designers may want to use.