[STIXTwoTextVF-Roman.vfj] Big difference in `b.f_` & `h.f_` masters would cause problems in extrapolation

[twardoch]

With just interpolation, the big difference in vertical placement of the topleft node between upright masters of b.f_ & h.f_ is not a problem, but with extrapolation will cause issues even if the entire design would be OK for some extrapolation:

[tiroj]

There are reasons why the OT variable font format is not designed to extrapolate beyond axis extrema.

The angle there has to change, and I don’t see a tidy way to do that which doesn’t involve one or more points travelling in ways that will cause problems if extrapolated. Let me know if you have a suggested architecture.

[tiroj]

I’ve restructured this for v2.13, so that there is less vertical movement, but at some point extrapolation is always going to explode things.

[twardoch]

Imagine how you'd solve that if you had a Black :) In the top part of b & h intended for ligatures, insert an extra node in the diagonal segment to keep the angle of the diagonal part that sticks out, but then move the lower node up.

[twardoch]

Something like this would be a clean solution, no?

[twardoch]

(Ofc I agree that at some point, it will :) )

[tiroj]

Yes, that's pretty much what I ended up doing. There is still a fair amount of movement, since the graphical relationship of the top of the ascender to the hook of the f changes so much as the weight increases.