On 04/04/2022 09.38, Corentin via Core wrote:
> Hello folks.
> I'm currently trying to implement a conforming version of P2071.
>
> It's somewhat...difficult, as 10646 is not the most developer friendly source of truth.
>
> Are implementers expected to manually comb the code charts in the pdf of ISO 10646 to find the relevant names?
>
> I understand the desire to refer to ISO 10646 but...could we please add a notes along the lines of
> "The names and aliases are derived from the character names in UnicodeData.txt and the aliases classified as alternate, control or correction of NameAliases.txt"

"derived from" isn't helpful, I guess, because it doesn't say
"identical to". The latter is what would make the job of
implementers easier.

In any case, we should refrain from attempting to add words to
C++ that could interpret and possibly contradict that other
standard. What if some ISO 10646 name differs from UnicodeData.txt?
That is hopefully not a bug in C++.

> And if there are exceptions to that, I suppose we should find out?

Why? We refer to normative statements of another standard, and
the terms "associated character name" and "character name alias"
are clearly defined there.

If you believe the other standard is hard to use or is internally
inconsistent, maybe you should consider filing a Defect Report
against that other standard with a suggestion for improvement.

Jens