Quoting John Cowan <[log in to unmask]>:

> Andreas Johansson scripsit:
> > I've, BTW, always been somewhat mystified that someone ever came on the
> > idea of having graphemic zero indicate /a/ (or /O/ and so on depending
> > on language) rather than phonemic zero. In a language like Sanskrit
> > it may perhaps save typing, but it's certainly the last idea I would
> > have stumbled on. People are weird.
> Well, Tengwar (as used to spell Quenya, anyhow) is an abjad with mandatory
> vowel marks, but the step from that to an abugida is small, as JRRT
> himself indicates rather offhandedly in a footnote:
>         In Quenya in which _a_ was very frequent, its vowel sign
>         was often omitted altogether.  Thus for _calma_ 'lamp' _clm_
>         could be written.  This would naturally be read as _calma_,
>         since _cl_ was not in Quenya a possible initial combination,
>         and _m_ never occurred finally.  A possible reading was
>         _calama_, but no such word existed.

What he failed to mention was that initial vowels go on a zero consonant known
as a "short carrier". Presumably, this would be retained when the _a_
diacritics were dropped, which would explain how one can know that **_acalma_
or **_acalma_ wasn't intended.

> And indeed we see that the Ethiopic and Brahmi abugidas descend from
> an abjad.  Canadian Syllabics is also an abugida, where the vowel
> signs are rotations and the virama (that's the word someone was
> looking for as a replacement for "vowel killer") is superscripting.

That was me. Thanks!