[Harfbuzz-indic] Unicode vs OpenType: nukta after halant

[Jonathan Kew]

On 18 Jul 2011, at 19:58, Bernard Massot wrote:

> Hello,
> 
> Another point on which Microsoft's OpenType specification disagrees
> with Unicode specification: OT says nukta can follow a halant (or a
> vedic sign) in Devanagari, whereas Unicode says nukta must appear before
> a halant.
> 
> OT states "Reorder marks to canonical order: Adjacent nukta and halant
> or nukta and vedic sign are always repositioned if necessary, so that
> the nukta is first.". By the way OT's specification is
> self-contradicting since syllable regular expressions don't allow that.
> 
> On the other hand, in chapter 9.1 of Unicode standard, Devanagari
> rendering rule R9 states "The nukta sign is placed immediately after the
> consonant in the memory representation. If the consonant represents a
> dead consonant, then NUKTA should precede VIRAMA in the memory
> representation.".
> 
> I seems clear to me that when a consonant both has a nukta and a halant,
> nukta must come first. If a nukta follows a halant, a dotted circle
> glyph should be inserted before it.

No.

Although the order <consonant, nukta, halant> is logical, and chapter 9.1 clearly advises this ordering, the two sequences <cons, nukta, halant> and <cons, halant, nukta> are defined to be _canonically equivalent_ in Unicode. Nukta and halant have distinct, non-zero combining class values, and therefore normalization will regularize the order of the marks.

Because the two orderings are canonically equivalent, they should both be expected to behave (including display) identically.

JK