Typo in Smith joint shadowing-masking function about gltf HOT 2 CLOSED

In Heitz (2014), the separable masking-shadow function is given as
$$\dfrac{\chi^+(\omega_o\cdot \omega_m)}{1+\Lambda(\omega_o)}\dfrac{\chi^+(\omega_i\cdot \omega_m)}{1+\Lambda(\omega_i)}$$
where $\omega_m$ is the microfacet normal. You're rightfully stating that it vanishes when $L$ or $V$ are facing away from the microfacet.
But, in the glTF spec Heitz's $\omega_m$ corresponds to the half-vector $H$, not the macrosurface normal $N$. So the current definition of G seems to be in line with Heitz. Consequently, $\chi^+(V\cdot N)\chi^+(V\cdot N)$ would check the macrosurface visibility.

With the current definition of $H$, these heavyside terms are indeed not particular useful. But, considering that some renderers may use sampling methods to generate $H$ this check becomes important again. Maybe we should add this info as clarification of why these terms are there.

from gltf.

@TheRealLL4 Does that information clarify this enough? Rather than edit the specification to add a note, perhaps we can point folks to this issue if they need the clarification.

from gltf.