stacked vs. side-by-side conjuncts

Peter Constable

Just wondering out loud: In the early days of OpenType, was it ever suggested to register a feature (or pair of features) for Devanagari or other Indic scripts to select between stacking vs. side-by-side conjunct forms? (Analogous to features for, e.g., lining vs. old style figures.) If not, why not?

John Hudson

How do you imagine such features being activated? How would they interract with Unicode control character sequences that form either half-form horizontal or explicit halant conjuncts? Since some conjuncts only take either vertical or explicit halant (no half-form) how would these features interract with reph and ikar reordering?

The latter is the trickiest problem, because the Unicode control character model strongly presumes that traditional ligature forms are the default if they are present in a font, i.e. the formatting control characters can be used to prevent ligature formation in various ways, but there is no control character method to force ligatures if some other shaping were default. One can imagine glyph level ligature formation such that e.g. horizontal half-form or explicit halant could be the default, and then traditional vertical stacks could be handled in e.g. dlig, but for the reph and ikar reordering issue, since dlig and all similar discretionary features are applied after reordering.

This is what prompted me to suggest, a while ago, the idea of adding a pre-reordering discretionary conjunct alternates feature.

John Hudson

[Another, related idea I have had is a software level solution that would work with input, editing, and display of Indic text. The idea is to have text ‘modes’, in which ZWJ and/or ZWNJ would be algorithmically inserted in a text to display it according to various preferences. This enables display of text to be controlled by the reader rather than by the document creator, and does so in a way that is easier and more consistent than having an author manually add control characters.]

Peter Constable

I wasn't necessarily thinking of adding this for Devanagari. In hindsight, if it were being encoded from scratch today, it might be done rather differently. (Separate visible killer from conjunct forming control; reph and eyelash ra.)

It just came to mind while I was reading a proposal for Tulu-Tigalari, and stacked vs. side-by-side is relevant but left as a font-level issue. I don't think it has any of the issues you raised.

John Hudson

Separate visible killer from conjunct forming control

I may be alone in this, but I don’t think it is too late to add an explicit visible virama to all the scripts that are missing it.