On 06/05/2020 01.12, Steve Downey wrote:
> Precomposed characters are the norm only in Western alphabets, where the combinations are few, and there are historical mappings for round trip purposes. In other languages combining characters are the norm, and there do not exist precombined characters at least for the most part.
>
>
> Devanagari, for example, has combining marks as the norm. That we can have a z with umlaut, which doesn't exist in natural language, falls out from the general mechanism, but we don't want to preclude the general mechanism.

Thanks; I wasn't aware of that.

> However, accomplish that by token pasting is a non-goal.

It seems we're on the same page, and also with our understanding
that lone combining marks (which are not in XID_Start)
are ill-formed even during translation phase 4.

Jens