mappings-01 and the general procedure

[John C Klensin]

--On Sunday, July 12, 2009 13:50 -0700 Erik van der Poel
<erikv at google.com> wrote:

>...
> Also, I would prefer an IDNAbis that does not specify applying
> NFC twice (once in the mappings draft and once in the protocol
> draft). It doesn't seem necessary.

I don't know if this is helpful, but Protocol does not specify
"applying NFC".   It only specifies (in Section 5.2) that the
string MUST be in NFC form before being processed further.   It
would be reasonable for the notorious section 5.3 of Protocol to
contain an explicit statement that the result of any
preprocessing or mapping step must not produce a result that is
not (still) in NFC form.  I think that is implied, but it might
not be clear and is relevant whether the Mapping draft is
retained and improved or dropped, so I'll add it to Protocol-13.


So, if the mapping document specifies NFC as a last step, the
language in Protocol would not require doing it again.   If the
mapping document does not specify NFC as a last step, the
Protocol requirement would essentially demand that NFC be
applied (or at least verified which, as Mark has pointed out, is
a low-impact and low-cost operation) before the string is
processed further.

My sense is therefore that, while the text should probably be
improved in Protocol, Mapping, or both, we really do not have a
problem here.

   john