SuttaCentral
Yes, that’s what I was trying to say. When I started learning this stuff I had (and still use) a chanting book that uses ṁ, and it seemed logical for the over-dot to be nasal and the under-dot to be retroflex. ṃ breaks that logic.
Unfortunately, since the modern publications seem to use ṃ, I guess we have to conform…
For me the vile abomination, or at least absurdity, is the use of digraphs —kh, gh, ch, jh, etc. — to represent single aspirated consonants. Some years ago I set about devising a better romanising system that would eliminate these. In place of the digraphs I mostly used the characters that the IPA used to use for implosives; on account of the great rarity of these, the said characters have now been removed from the IPA, obviating any risk of confusion. While engaged in this it occurred to me that I could also free the alphabet of Ven. Sujāto’s abomination, the underdotted m, by the simple expedient of consigning all dots — gruesome, ghastly things! — to the outer darkness.
This was the result:
Oh, that’s nice, congratulations. It’s much closer to the Indic orthography, more precise but still very readable.
Are those characters all found in Unicode?
Yes, they’re all there.
I see now that I mistakenly posted an earlier prototype. In the final one I made changes to the retroflex consonants because the letter eth looked completely out of place in the earlier one.
sara: a ā i ī u ū e o
kaɳȶaja: k ƙ ɡ ɠ ŋ
tāluja: c ƈ j ʝ ɲ
mudɗaja: ʈ ȶ ɖ ȡ ɳ
dantaja: t ƭ d ɗ n
oʈȶaja: p ƥ b ɓ m
avaɡɡa: y r l v s h ɭ
niɡɡahīta: ã ĩ ũ
The letter g, by the way, is not the normal one but the “Latin small letter script G”, found in the Unicode IPA extensions: ɡ. It’s more elegant to use this because it forms the basis of the “Latin small letter G with hook” (ɠ), which I use in place of gh.
There is a small nuance here: ṃ stands not only for pure nasal but also for a nasalized sonorant like w̃, otherwise it would be rather difficult to explain why you cannot use *āṃ, *īṃ, *ūṃ. As I said before,it is not an easy task to find out which phonological value the anusvara character originally had in which word. Besides, if we assume that ṃ always stands for pure nasalization, it becomes extremely hard to explain such forms as upasaṅkamati. For these reasons, I would retain the anusvara character in the reformed Pali Romanization, but maybe we should look for a more appropriate character.
I don’t assume that it is always a pure nasal, only that this is its intrinsic sound, while the other ways in which it may be realized are derivative.
What I was taught is that ṃ is realized as a pure nasal either when it’s followed by any avagga consonant except y, or when it occurs as the final consonant of a word. When followed by a vagga consonant or by y it’s a homorganic nasal whose value is in each case determined by the vagga of the consonant that it precedes. And so one writes (or at least may write) kathaṃkathā, evaṃca, yaṃṭhānappatto, kiṃti, evaṃnāmo, taṃpi, evaṃ me, saṃyojana, but one pronounces them (or ought to pronounce them) kathaṅkathā, evañca, yaṇṭhānappatto, kinti, evannāmo, tampi, evamme, saññojana.
That is my point. It can hardly have a pure nasal as its intrinsic phonological value if it has allophones in consonantal environments. It just doesn’t happen to pure nasalized vowels, or it may happen to them in some obscure languages but I am not aware of it. For a nasal sound to assume other phonological values like n, m, or ñ, it has to be a real sonorant, not a mere nasalization, i.e. -aṃ is realized as two sounds, not one. In our case, ṃ most likely was a nasalized miture of w and m. However, quite often the ṃ character did stand for pure nasalization at the word boundary. Since the intrinsic value is most likely to be a sonorant, and it is a hard task to figure out where ṃ stood for nasalization, it is best to stick to a separate anusvara character.
This delightful nerdiness should probably be split into it’s own thread.
I just went hunting for this thread to try and understand why I was seeing memerise showing upasaṁkamati in the AK Warder stack, yet upasaṃkamati in the book (printed by PTS in 2005), and what the difference is.
Now I have a headache 
I will take a break and re-read.
