The math doesn't work well

[robmck-ms]

As mentioned in the README, the math for this the proposal may be problematic: First, the proposal increases the level of interdependence between axes, which, when combined with already mutually dependent axes like weight, width, and optical size, could easily create regions that are even more tangled than without XVAR. Second, XVAR could add non-linearities and regions of instability. Thus, XVAR may cause more problems than it solves.

TODO: add more detail here

[davelab6]

@be5invis do you think your zvar proposal has preemptively addressed these concerns?

[LettError]

Can that be read somewhere?

[davelab6]

+1

[thundernixon]

Looks like it might be here: https://github.com/be5invis/OTVAR-ZVAR-Table

I don't know how much more there is to the ZVAR proposal than in that readme, but if it also solves problems by allowing for non-linearity, doesn't that almost necessarily mean that complexity would be added, having the same potential issues as XVAR? Further, despite rising complexity introducing new potential problems, is that necessarily a bad thing? (Note: I am not a mathematician, just a designer hoping for the possibility of non-linear control in my fonts).

[be5invis]

@robmck-ms From my aspect you can turn dependent axes (wght, opsz, etc.) into independent parametric axes using XVAR (or ZVAR), which could be a solution to the dependency problem. Adding another layer of indirection (XVAR, ZVAR, AVAR, ...) would always make the equation more complex.

[davelab6]

Looks like it might be here: be5invis/OTVAR-ZVAR-Table

Alas, ZVAR and XVAR are kinda competing proposals

hoping for the possibility of non-linear control in my fonts

Does AVAR already deliver that control for you? :)

[behdad]

I still think what we need is avar v 2.0

[LettError]

From my aspect you can turn dependent axes (wght, opsz, etc.) into independent parametric axes using XVAR (or ZVAR), which could be a solution to the dependency problem.

Are you sure about that? Independent parametric axes are a great idea but incredibly difficult to turn into practice, into typefaces where they make sense.

[thundernixon]

@davelab6

Does AVAR already deliver that control for you? :)

It definitely helps, but as far as I understand it (albeit, not completely), AVAR doesn't give the control that would be needed to warp designspaces in a few useful ways. Some examples I'm thinking of now are:

• in a 2D weight & width family, warping weight differently for different widths (say, to fill the corner of a trapezoidal designspace without duplicating a master (increasing file size) or recomputing all input values (potentially messing with sources when it would be nice to do so in mastering)
• allowing for other tricks with duplicated masters like higher-order interpolation, as described by Underware Type Foundry (including this in case others come across this who aren't familiar with it yet), without increasing final file size, and possibly with the ability to put this into a single axis control (not sure if that stretches beyond these two proposals)

I can't provide a specific example right now, but I suspect that complex families like Knockout, in which lettershapes change between widths and weights, might require more control than AVAR delivers to exist in reasonable file sizes.

P.S.: I hope my earlier statement "I am not a mathematician, but..." doesn't come off as being dismissive to anyone – I very much appreciate that there are people here who are figuring out the mathematical complexities of this field and working to keep things from going off the rails.

@behdad Are you suggesting that some of these ideas be worked into the next version of avar, or is there already another proposal or thread related to it?

[behdad]

@behdad Are you suggesting that some of these ideas be worked into the next version of avar, or is there already another proposal or thread related to it?

The former.