I just realized I misunderstood what your draft was saying:
> Recommended order of applying canonicalization rules is as follows:
>
> (1) "idntabjpcanon10"
Despite its name, this file has COMPATibility mappings.
> (2) "idntabjpcom10" [presumably "idntabjpcomp10" was meant]
And this one has CANONical mappings. The file names are misleading
given existing terminology, and I was thinking they were the
other way around.
Now I think you're saying the client should fold compatibility
characters (full- and half-width variants) and the resolver
accept any valid Unicode spellings, including the nonspacing
forms of the voiced and semi-voiced marks, but not
compatibility characters. And that it doesn't matter to
Japanese users if clients written for other languages don't
fold the width variants. I agree with all of that.
