L2/11-351 Subject: RE: Grantha sub-base vocalic L/LL and Vedic Tone Asterisk Above Date: Mon, 29 Nov 2010 23:34:24 +0000 From: Peter Constable There are some cases in the previously-encoded Indic scripts in which certain elements can occur in different positions. For instance, IIRC the medial ra (rakar) in Telegu is sometimes written below the nominal consonant, sometimes to the sometimes to the left, and sometimes (less common?) to the right. I even have a Telegu newspaper that has a medial ra in their name, and it appears one way (below) in the masthead on the front page and a different way (left) in the same word when in headings. The usage I've seen suggests that this correlates with the font; that is, I haven't seen both forms used within a body of text formatted in a single font, or even in a single font applied to different pieces of content. For sake of discussion here, I'll assume that the Grantha usage is comparable in these regards to what I have seen for the Telegu case I've described. The usage suggests to me that there isn't a usage scenario requiring encoding as separate characters. Put another way, I don't know of any need to differentiate the positional variants in plain text. It appears that a single encoded character is adequate in terms of encoded representation of information even though different presentations are possible. That said, supporting such positional variation in rendering implementations can be tricky. If rendering implementations required fonts always to handle consonant + modifier mark combinations using ligature glyphs, then there wouldn't be any issue. But OpenType is generally intended to give alternatives to the font developer as to how they implement their font, and one of those options is to have separate mark glyphs that can be positioned relative to the base. But for that to work, the rendering implementation must know whether a modifier character is to be presented to the right of its base, to the left, or above/below: in terms of OpenType processing, - a glyph presented on the right is processed like any base glyph - a glyph presented on the left is processed like a base glyph, but the rendering engine must re-order glyphs to place it in the output before the base consonant - a glyph presented above or below is processed like a mark glyph, not a base, and when mark placement is done it must be located in the glyph sequence directly after the base to which it must anchor; in some cases, this may require some re-ordering in the glyph sequence (e.g., in Odia and Kannada scripts, an ii matra may need to be re- ordered before a combination so that the matra is positioned above the base consonant not the modifier "vattu" consonant) For the Telegu case, we implemented support in our rendering engine for different font-determined positioning of the rakar, but it's quite tricky, and requires that we run heuristics on the font to assess which behaviour it assumes. For these reasons, it would be much easier for rendering implementations if these positional variants were encoded as separate characters. This comes down to a trade-off between complexity in rendering implementations versus the usability issues raised by having separate encoded characters that represent the same "thing" (as the user perceives). Peter -----Original Message----- From: Shriramana Sharma Sent: Wednesday, November 24, 2010 7:47 AM To: Peter Constable Cc: Deborah Anderson Subject: Grantha sub-base vocalic L/LL and Vedic Tone Asterisk Above Dear Mr Constable, I did not have the opportunity of meeting you (via phone) at the last UTC Meeting because (apparently) you weren't able to attend (at least during the discussions that *I* attended). The two issues of Grantha sub-base vowel signs for vocalic L/LL (L2/10-341) and Vedic Tone Asterisk Above (L2/10-349 and L2/10-398) were deferred at the meeting pending your comments on the same. As I am the author of the above proposal documents, I would like to know what feedback you had/have to give on the above documents -- whether these two characters: 1137E GRANTHA VOWEL SIGN SUB-BASE VOCALIC L 1137F GRANTHA VOWEL SIGN SUB-BASE VOCALIC LL and this one character: 1CF7 VEDIC TONE ASTERISK ABOVE are valid candidates for distinct encoding or not, in your opinion. Thanking you, Shriramana Sharma.