- From: Richard Ishida <ishida@w3.org>
- Date: Fri, 11 Jul 2014 22:11:04 +0100
- To: John Hudson <tiro@tiro.com>, public-i18n-indic@w3.org
On 11/07/2014 20:52, John Hudson wrote: > On 25/04/14 9:32 AM, Richard Ishida wrote: > >> If a conjunct doesn't display the virama, it's components and associated >> vowel-signs, diacritics, etc should be kept as a single unit, but >> otherwise the top bar is broken around each akshara. For example, one >> would expect to see अंतर्राष्ट्रीयकरण [1] stretched as >> >> अं त र्रा ष्ट्री य क र ण >> >> If, however, the virama is displayed explicitly, one would expect to see >> the same word stretched as >> >> अं त र् रा ष् ट् री य क र ण > > Unless an explicit virama has been triggered by insertion of ZWNJ (see > TUS fig9.3), I don't see any 100% reliable way to determine whether a > particular conjunct in a particular font displays with an explicit > virama or not. For an OpenType font using <dev2> shaping, it might be > possible to get an indication from querying the glyph string after GSUB > cluster shaping features have been applied -- presuming that the font > lookups have been made in a predictable way -- looking for glyphs that > map to the virama character. But for the older <deva> shaping it is > common for fonts to contain nominal half form glyphs (accessed via the > <half> feature) that include a visible virama (below retroflex letters > such as ट). In that case you have a visibly explicit virama but nothing > in the glyph string to indicate that it is there. Nor can you make an > assumption that a specific subset of letters will be handled in this > way, because छ may display with explicit virama or with a true half form > depending on the individual typeface design. Good points, but my question at this point, rather than worrying about the technical difficulties involved, revolves around what indic typographers, printers, designers and users actually expect to see. If an akshara is always a unit, whether or not the virama is visible, then i think the technical solution is more straighforward. But so far I'm not hearing what the actual requirement is. RI
Received on Friday, 11 July 2014 21:11:34 UTC