[HarfBuzz] Rendering Manually Line-Broken Indic Text

[Richard Wordingham]

Indic text is not always broken at akshara boundaries.  What should a
HarfBuzz client do if it wishes to render such text?  What are the
support requirements from the transcriber and a font that can already
cope with text only broken at indisputable akshara boundaries?

The reason I ask is that I am trying to preserve line boundaries when
transcribing some Tai Tham text written on palm leaf, along with notes
in English, in a rich text environment. The application I am using is
LibreOffice; it may be that this task is beyond its intended
capabilities.  If I raise a feature request on LibreOffice, it may help
to give some indications of how HarfBuzz should be used in this special
case.

The text I am transcribing is overwhelmingly in Pali and is
written using scriptio continua.  There is almost no punctuation.  The
writer did not comply with Everson's diktat that line-breaking is
lexical - it is not even broken at akshara boundaries, but rather at
column boundaries. (This was normal practice.)  Thus I get the following
rendering issues:

1) A line may be broken between a preposed vowel (being Pali, this is
always U+1A6E TAI THAM VOWEL SIGN E) and the rest of the akshara.

2) A line may be broken within an akshara between the consonant stack
and a trailing vowel.  So far this vowel has always been U+1A63 TAI
THAM VOWEL SIGN AA; I suspect U+1A64 TAI THAM VOWEL SIGN TALL AA would
not make sense at the start of a line.

I expect to encounter the following similar, additional issue:

3) A line may be broken between U+1A58 TAI THAM SIGN MAI KANG LAI at
the end of an akshara and the next akshara, for in this text U+1A58
behaves not as Consonant_Final but as Consonant_Preceding_Repha.

How should the resultant dotted circles be suppressed?  I do not wish
their suppression to introduce spaces.

I would prefer the font to be compatible with the native renderers of Windows 10, MacOs and iOS.

Richard.