No subject

have to support the "correct" IRI to URI transformation,
the variant based on IDNA for host names.

Once they got that it is obvious to use exactly the same
recipe (=3D pure IDNA) also for bare host names outside of
IRIs.

But if they pick the "wrong" transformation, where host
names end up as percent-encoded UTF-8 non-LDH gibberish,
we get an interoperability nightmare. =20

3987bis "MUST NOT" permit the wrong transformation.  But
this mustard is IMO a job for 3987bis, not for IDNAbis.

 Frank




=20