https://github.com/harfbuzz/harfbuzz/issues/2017 should provide the context for this.
Ever since the early days of Devanagari Unicode, scholars like me dealing with Vedic Sanskrit orthography have been experiencing this problem, but chalked it upto early days and consequent insufficient support for Vedic sequences. Even now, Vedic support even on the font side is quite limited, and we also find limitations on the software side. So I hope it's time to fix them one by one. The issue I would like to discuss now is as follows: # SEMANTIC DISSOCIATION OF THE VISARGA FROM THE SYLLABLE In Vedic, syllables that carry tone markers – which are mostly above-base or below-base – often have to take a visarga, which is always post-base. In this case, the sequence intuitive to native scholars like me is: <C*V SYLLABLE> + <TONE MARKER> + <VISARGA> This is because the tone marker indicates the tone of the syllable (or its vowel) and the visarga is a separate aspirated sound *after* the syllable to which the tone marker doesn't apply. In fact, the only reason the visarga sign is analysed as a combining mark rather than a separate letter is that it is not used in isolation without a preceding syllable. Otherwise ie linguistically it doesn't modify the preceding syllable in any way. Anyhow, the point is that the tone marker should come before the visarga because it semantically applies to the preceding syllable and not the visarga. This is all the more so since in some Vedic contexts (Sama Gana) the visarga is far separated from the syllable by other syllables like digits (themselves carrying combining marks) or spacing anusvara, as seen in examples from my Grantha proposal L2/09-372 p 40. So the visarga is semantically quite dissociated from the preceding syllable unlikely the tone marker which is intimately associated with it. # SAME APPLICABLE TO THE ANUSVARA The same argument is also applicable to the anusvara as it also represents a nasal sound separate from the preceding syllable. (The candrabindu OTOH nasalises the preceding syllable itself.) The above Grantha proposal page also shows an example where an anusvara is orthographically separated from the preceding syllable by three characters: a tone marker + avagraha + digit. L2/15-178 shows that in equivalent contexts of Devanagari the digit 0 is used as a substitute since the Devanagari anusvara is non-spacing. All this goes to the dissociation from the syllable of the anusvara – just like the visarga – compared to tone markers. So to be consistent, even in case of Devanagari (or such script) where the anusvara is non-spacing, the sequence when a tone marker is also involved puts the tone marker first, as mentioned before: <SYLLABLE> + <TONE MARKER> + <VISARGA/ANUSVARA> # CURRENT SITUATION INCOMPATIBLE WITH ABOVE However, even the simplest Vedic sequence (not involving Sama Vedic or multiple tone marker combinations) like दे॒वेभ्य॑ः throws up a dotted circle, and one is expected (see developer feedback in that bug report) to input the visarga before tone markers, hoping the software is intelligent enough to skip over the visarga (or anusvara) place the tone marker over the preceding syllable correctly. Why it is necessary to put the visarga first in input only to have to skip over it in shaping is beyond me. So makes sense neither from a linguistic nor technological perspective to push the tone markers to the end of the syllable. Even the developers acknowledge that non-spacing marks are normally (ie outside Indic) input before spacing ones. However, they say “we can't support that in this particular case because this is how Microsoft does it and we have to follow suit to ensure people get the same shaping for the same input”, notwithstanding the fact that the expectation to put the visarga/anusvara first is non-sensical as explained above. So everyone is looking to Microsoft Uniscribe (or whatever its successor is) to fix things first before they can follow. I figured that if this is discussed and decided here, everyone can fix it at the same time. -- Shriramana Sharma ஶ்ரீரமணஶர்மா श्रीरमणशर्मा 𑀰𑁆𑀭𑀻𑀭𑀫𑀡𑀰𑀭𑁆𑀫𑀸
