n8willis
commented
4 years ago So, looking back on this now, I suspect that my sole reason for choosing "dda" for the illustration was that it was the only nukt lookup in Noto Gurmukhi for which it was visually clear that the application itself was a substitution with a whole new glyph and not a repositioning for the mark.
I definitely tried to choose illustrations that were more resistant to that sort of misunderstanding where there's a possibility of doing so. Still, that is not as important of a factor as picking a typographically correct example. I glanced at Noto Sans & Serif again and didn't find any candidates that would check-off both the "correct" and "visually not a GPOS move" columns, so I'm totally happy to make the change.
But, as a general principle as well as here, if there's something that I've missed that would also help fill that visual-unmistakability goal, please let me know. (I feel like examples in shaper docs can lean more towards "always be hard to misunderstand" whereas examples in docs for type designers need to lean more towards "most useful in practice", but that's not a scientific theorem or anything.)
https://github.com/n8willis/opentype-shaping-documents/blob/master/opentype-shaping-gurmukhi.md#32-nukt
The “ਡ ਼ → ੜ” substitution is a widespread but inappropriate usage of this feature. Please use a different example. Not sure if this information is valuable to shaper development, but font producers should be aware of this bad practice, so I’ll make sure to discuss this in http://github.com/typotheque/text-shaping.
I recommend “ਸ ਼ → ਸ਼” as the example, because ਸ਼ is both a frequently used nukta-ed letter and the first one in the alphabetical list of nukta-ed letters: ਸ਼ ਖ਼ ਗ਼ ਜ਼ ਫ਼ (ਲ਼)