Charter changes and a possible new direction

[Michel SUIGNARD]

Paul, I have a lot of sympathy for your approach. I made similar efforts
a while ago to upgrade stringprep and consequently nameprep to
accommodate obvious flaws in bidi and Unicode coverage but without
changing the overall model concerning mapping and unassigned characters.
I thought it was a reasonably approach that would not scare the software
implementers (unlike frankly the new approach which at least on first
approach can be perceived as a significant incompatible update).
Obviously I was not too successful.

At the same time we have to recognize that additional issues have
surfaced in bidi and contextual rules (concerning some Indian languages)
that require significant changes in the algorithm. So whatever the
approach, these have to be accommodated.

I had also hoped to decouple the Unicode version dependency by pushing
the relevant table in an IANA maintained set which didn't require a new
RFC for new Unicode versions. At the same time, now with IDN syncing
with Unicode 5.1 (or even 5.2 if we wait another few months), the
impetus to sync with newer versions will fade because the added
characters are becoming less and less significant. Unicode/10646 as a
repertoire is maturing. I should know. So embedding the Unicode data in
the RFC is not as serious an issue as it was when IDNA2003 was created.
I however favor the usage of external properties referencing because it
reduces the chance of errors.

I don't have a clear opinion of the chartering issue, but share Martin
Duerst opinion that I don't see that a big deal.

I think that getting something done quickly that can be endorsed by all
constituencies, including major implementers of existing IDNA2003
libraries, is the key priority. My gut's feeling is that your approach
could achieve that.

Michel