Transliteration of variant anusvāras

[danbalogh]

This is a TG revision issue.
We currently distinguish the following members of the anusvāra family:

• anusvāra, transliterated ṁ
• anunāsika/candrabindu, m̐
• Cam anusvāra-candra, m̃
• Javanese/Balinese ulu ricem, ṁ*
• Bengali alternative anusvāra, ṁ*

Main question, primarily to @arlogriffiths: given that the latter three variants are specific to languages/regions and are extremely unlikely to co-occur, is there any reason why we should not transliterate them in the same way? This would also provide for other anusvāra alternatives should the need ever arise, e.g. Gurmukhi ṭippī.
I would be happiest if we could forget about the asterisk and just use m̃ to transliterate all instances of "region/language-specific variant anusvāra". If the Cam anusvāra-candra is so different in function from the other two that a distinction seems essential, I would still prefer to avoid the asterisk and propose that we transliterate all other anusvāra variants using m with a different diacritic.
Most straightforwardly, all of the alternative anusvāras could be transliterated ṃ. This is slightly in conflict with using ṃ for Pyu underdots, but I don't suppose a conflict can arise, and at any rate, it may be a good idea to use something else for the Pyu underdot if possible. Otherwise, m̃ is also fine by me (if the Cam variant does not have to be kept apart from the others), or we could use ɱ (available as a precomposed unicode character and apparently used in some contexts for transcribing pre-nasalisation, so not completely alien from the anusvāra).

Secondary question: do we have anything like a clear idea of the function of some of these variants? Anunāsika is pretty straightforward and ulu ricem is already stated in the TG to represent a syllable-final /m/. But (@arlogriffiths, @salomepichon) what does an anusvāra-candra do when it's at home? And is there any systematicity in the Bengali alternative anusvāra (@ryosukefurui )?

Bonus question: can anyone provide an image or two of the following variants:

• anunāsika/candrabindu (from an inscription or, if not available, from a manuscript)
• anusvāra-candra
• ulu ricem (something clearer than the image now in the TG would be nice)

[arlogriffiths]

MAIN
I understand all your arguments and of course can see their force. Nevertheless, I'd prefer to stick with what we have, not only because the various choices reflected in our TG in the versions released so far reflect often long processes of decision making in the respective fields, but also because all the different members of the anusvāra family in question have different graphic shapes as well as functional ranges. It would seem inconsistent to me to expect use of ṁ for (what I perceive to be a fairly specific family, namely) all cases of dot or small circle atop a consonant character, while also expecting use of the single representation with a single transliteration device of things so graphically diverse as anunāsika/candrabindu, anusvāra-candra and ulu ricem. Anyhow, I believe ISO 15919 has a specific recommendation for candrabindu and I'd prefer to stick with it.

I would certainly not advise generalizing m̃ beyond its use for anusvāra-candra, because it is hard to type for many people and also poorly displayed in many fonts. If any character had to be generalized, ṃ or ɱ would be better candidates. I agree that in due course we probably something other than ṃ for the Pyu underdot, but no suitable option has emerged so far. As long as we don't have an alternative, I think the risk is too great that a Pyu Skt. inscription with candrabindu will emerge that would lead to ambiguity of ṃ if it were adopted also for candrabindu.

SECONDARY
In Cham-language texts, anusvāra-candra almost always functions as equivalent to m· (so /m/). As the example below shows, in Sanskrit context is can function as equivalent of anunāsika/candrabindu.

BONUS

• example of anunāsika/candrabindu: Om̐ namaś śivāya from DHARMA_INSCIC00141

• example of anusvāra-candra: tad-vam̃śe from DHARMA_INSCIC00252

• example of ulu ricem: dharmmaś ca janāti nara ya vr̥ttaṁ* from DHARMA_INSIDENKSarwadharma

[danbalogh]

Thanks, but you seem to have missed an important detail. I never suggested changing the transliteration of anunāsika to anything other than m̐. My proposal of a uniform transliteration applies to "the latter three variants [which] are specific to languages/regions and are extremely unlikely to co-occur". [BTW, in the new TG, also I want to propose the same transliteration for any other regional anusvāra variants that do not occur in our texts but might in others, such as the ṭippī in Gurmukhi and possibly the Telugu half-anusvāra.] Hence your observations about the ISO-15919 recommendation for anusvāra and the possible occurrence of candrabindu in a Pyu inscription are beside the point.

As for graphic diversity, I don't think it should be relevant. For one, ṁ already transliterates not just a dot or small circle above the character, but also the Telugu anusvāra ం, which is a large circle in its own segmental space after the character. Next, our current use of ṁ* also already transliterates two very different-looking signs (ulu ricem and Bengali anusvāra). Further, graphic diversity should not matter anyway, just as it does not in the case of regular "letters" (where e.g. t transliterates Brāhmī, Siddhamātrikā, Tamil and Khmer, etc. ). The only legitimate reason I see for keeping separate transliterations is that there is (or may be) more graphematic difference between these anusvāra relatives than between regional allographs of the same original Indic grapheme. But if we wanted to use separate transliterations for these, then by the same logic separate transliterations will also be needed for any additional variants not in our present scope, and I don't want that.

All in all, it seems to me that you would be amenable to getting rid of the asterisk (for ulu ricem and the Bengali anusvāra), which I would very much like to do because the asterisk can be ambiguous, and especially because I'd like to avoid using two target characters for a single source grapheme (everywhere except where inherited from ISO, i.e. aspirates and diphthongs). It further seems to me that you're OK with continuing to use identical transliteration for ulu ricem and the Bengali anusvāra, and might even accept using the same transliteration for these two and anusvāra-candra. So if we can agree on that much, all we need to figure out is how to take it from there:

A. Which transliterations did you have in mind when you referred to "long processes of decision making in the respective fields"? It probably does not apply to the ulu ricem and the Bengali anusvāra, since you appear to be at least somewhat OK with getting rid of the asterisk.

1. If this applies only to the anunāsika, then no problem, I don't want to change that.
2. If this also applies to the Pyu underdot, then I can of course respect that, and don't insist on appropriating ṃ from that purpose.
3. And if it also applies to the anusvāra-candra, then that is also fine, we can keep the transliteration.

B. What should be the transliteration character for the two or three "regional anusvāras"? I see the following options:

1. ṃ. Advantage: relatively easy keyboard access, decent font support, makes intuitive sense to end users. Disadvantage: may be confusing to end users (who might think it's the same anusvāra as ṁ). More importantly: clash with Pyu underdot means we'll either have ambiguity (which is OK but which I'd prefer to avoid) or we'll have to arbitrarily change the transliteration of the Pyu underdot. [Note: the TG also mentions a Burmese underdot, currently transliterated in the same way.]
2. m̃. Advantage: least change to existing transliteration scheme, makes some intuitive sense. Disadvantage: difficult keyboard access and poor font support, but I'm not sure that this really is much of a problem: people who need these characters can always use private shorthand.
3. ɱ. Advantage: possibly better font support than for m̃ (I haven't checked, but at least there is a codepoint for the character, while m̃ has to be composed). Disadvantage: least intuitive (but still OK).

C. Should the uniform transliteration apply only ulu ricem and Bengali anusvāra (plus any future anusvāra spawn), or also the anusvāra-candra? The answer to this sort of depends on the answer to A and B: if it's important to keep the tilde for anusvāra-candra, but you are very averse to using the same for the other variants, then a new transliteration is needed only for the others; otherwise it would be most expedient to use the same transliteration for all of these.

All in all, my order of preference is the following:

1. Use m̃ for all regional anusvāra variants (anusvāra-candra, ulu ricem, Bengali anusvāra and any others that might be out there). We then get to keep the current transliteration of the Pyu underdot and the anusvāra-candra, and only need simply bulk replacement of m* to m̃, and we have this transliteration category of "region-specific anusvāra variant".
2. Use m̃ for the anusvāra-candra if this is a well-established convention, and revert to ṃ for all others, in which case we'd best change the Pyu underdot to something else, perhaps U+2022 • BULLET.
3. Use m̃ for the anusvāra-candra if this is a well-established convention, and adopt ɱ for all others, leaving the Pyu underdot alone.

Thanks also for the pictures.