May be this debate could find an end if there was a way to encode "private use variants", so that we can override an existing character with correct properties by creating a custom variant, which would immediately inherit the properties of the base character on which it is encoded.
But for now there's no private use variant codes (PUV). I think that a small block of 16 codes (may be even less) would be largely enough (given that it would be used only in pairs after any standard character). They could be used after any base character, possibly even after a combining character (so the default combining class for these PUV should be 0). For now there's still no way to have variant sequences unless they are registered and standardized by Unicode but registration should be not needed (forbidden) for sequences containing PUV. I think there's a usage pattern for such schemes. Their default (spacing) glyph could be a dotted circle with a single hex digit inside, it would be itself non-joining, it would be itself bidi-neutral and used only after a base character from which it would inherit the directionality (so the glyph would appear automatically on the correct side). Actual fonts implementing these PUV sequences would treat the PUV sequences as distinct unbreakable entities mapped to their own abstract character, and subject to common ligation. Le mer. 22 août 2018 à 04:58, Andrew Cunningham via Unicode < unicode@unicode.org> a écrit : > > > On Wednesday, 22 August 2018, Mark E. Shoulson via Unicode < > unicode@unicode.org> wrote: > >> On 08/21/2018 02:03 PM, Ken Whistler via Unicode wrote: >> >>> >>> >> Best we can do is shout loudly at OpenType tables and hope to cram in >> behavior (or at least appearance, which is more likely all we can get) that >> vaguely resembles what we're after. And that's not SO awful, given what >> we're dealing with. >> >>> >>> > At the moment I am looking at implementing three unencoded Arabic > characters in the PUA. > > For the foreseeable future OpenType is a non-starter, so I will look at > implementing them in Graphite tables in a font. > > Andrew > > > > -- > Andrew Cunningham > lang.supp...@gmail.com > > > >