--On Wednesday, 23 July, 2014 07:12 +0200 Patrik Fältström
<[email protected]> wrote:

> 
> On 22 jul 2014, at 20:27, Andrew Sullivan
> <[email protected]> wrote:
> 
>> On Tue, Jul 22, 2014 at 02:16:39PM -0400, Marc Blanchet wrote:
>>> - maybe it will be better to have different experts so that
>>> they look at the problem with different eyes (in this
>>> context, more eyes is a feature). but they would be better
>>> to coordinate. not sure it needs to be  written
>> 
>> I think it would be _terrible_ to have different experts
>> here, unless there's an interlock between then registries.
>> If there are two experts, they might draw different
>> conclusions.  Indeed, John, Patrik, and I all came to
>> different conclusions about the most recent version of
>> Unicode, and it was only after considerable discussion that
>> Patrik and I understood what John was worried about.
>> Different decisions between the experts would be extremely
>> bad.
>> 
>> I know there's a hurry to get this out, but hoping the
>> experts get it right if there's no interlock is too dangerous.
> 
> Being the appointed expert for IDNA, I would like to say I
> would LOVE to have a 2nd person for among other reasons the
> reasons Andrew list above.
> 
> That said, I think the appointed experts should get
> instructions to _communicate_with_each_other_ (!!!) and
> synchronize in a coordinated way with IANA. This to solve the
> issues I think Pete was thinking of addressing by having one
> and only one person.
> 
> I am expert on IDNA, and possibly one person with clue on
> Unicode.
> 
> Not Precis.
> 
> But with that knowledge, of course I am happy to continue to
> help IETF as much as I can.

Let me suggest that everyone quoted above is correct, but the
solution is a little bit different.

I think the solution, hinted at in Andrew's note, is that "the
expert"  be a team of several people, with different but
complementary knowledge, working together to conduct a review
for IDNA and PRECIS together.  They are finished only when they
agree, either that no changes are needed or on a single set of
changes.
  
We have now discovered a Unicode change issue that requires
careful human attention to detect and that cannot possibly be
detected by simple running of different implementations of a
test suite/ validation procedure.  When we set up the expert
procedure IDNA2008, we assumed that table-running with
independent implementations would be all that is necessary, but
that was based on assumptions that weren't true (sorry, but
anyone who wants to understand that case needs to read
draft-klensin-idna-5892upd-unicode70 -- I'd recommend waiting
for -01, which should show up sometime next week (possibly
sooner) with some additional clarifications).   Because the
algorithms are not sufficient (tghey are still necessary and
very important), the need for multiple people, with multiple
perspectives, to look carefully at new Unicode versions should
be clear.

I think experience from this round indicates that that expert
pool be at least three people (again, with different backgrounds
and perspectives).  Probably more than five would become
unwieldy.  

Because we seem to have disagreement on this subject, I believe
there is now a requirement that the "IANA Considerations"
section of "precis-framework" specify explicitly that:

        * There must be an expert team (or one expert and
        designated advisors), appointed as usual by the IESG,
        with at least three members, with diverse backgrounds
        and a recommended upper limit on membership of five or
        six.
        
        * That expert team is responsible for the reviews
        required by both PRECIS and IDNA, precisely because the
        base rules and categories, are expected/required  to be
        the same (note Peter's comment yesterday that, if the
        copies of the rules and categories used in PRECIS IDNA
        were different from those of IDNA, that was a bug that
        needed fixing).
        
        * The experts are finished, and IETF protocols ready to
        safely use a new version of Unicode, without risk of
        strings that are valid being disallowed, only when they
        agree (even if that agreement requires agreeing to
        disagree on details but move forward).

I'm happy to work with Peter on specific text for the document
if needed.

I can't stress how important this is.  If we don't have
agreement on it, some assumptions that were critical to
yesterday's agreement and consensus are false and the consensus
is meaningless.

If we have different experts with nearly-final authority for
different protocols, it takes us back into exactly the situation
we discussed yesterday -- the base exception lists, tables, and
allowable characters of IDNA and PRECIS will diverge (whether
they are the same today or not).  That is different from the
"profiles", which I would like to think of as per-protocol
exception lists from a base category and rule collection, being
different -- that latter is something we have to live with for
historical or specialty reasons.   If our intent is to design
things to allow that divergence to occur, then yesterday's
discussions about avoiding divergence for user confusion and
maintainability reasons and maybe the one about having a clear
interface for applications developers haven't had any effect,
perhaps suggesting that some of those who agreed to them didn't
understand them.

   regards,
     john



 (even, if PRECIS persists in establishing mappings, pro


_______________________________________________
precis mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/precis

Reply via email to