Strong recommendations to/from registries about character handling

[John C Klensin]

Hi.

This may just be lack of sleep, but I've noticed a commonality
among Vaggelis's note about Final Sigma, the note from Jaeyoun
Kim about Jamo, and some discussions occurring on other lists
about character treatment (above the protocol level) for Arabic
script.   Perhaps the JET recommendations about CJK fall into
that category as well.  This observation, together with the
discussions of the last few days, is changing my thinking on the
subject(s) somewhat.

In each case, the folks with day-to-day experience in the use of
the script on the Internet, having carefully examined its actual
or potential use in domain names, have analyses and strong
recommendations of the character of "if you are going to permit
registrations involving this script, you should pay careful
attention to this issue or you will effectively encourage bad
behavior".  

Most of those analyses point to decisions that are above the
protocol level and hence outside the scope of this WG, but that
does not make them less real.  Vaggelis expresses concern that,
while the Greek registry will understand the issue and do the
right thing, the gTLD registries and most other registries that
accept Greek-script registrations will not.  I think that is a
legitimate concern, even though I don't think it is sufficient
to justify our trying to push the recommendations into the
protocol level.

The question is what to do.  While some people's instinct is to
dump the problem on ICANN, ICANN has (as Eric pointed out
recently) very little practical authority: no relationship with
most ccTLDs that would give them enforcement capability, no
relationship at all with most registries for second level
domains and below.  It is even unclear how much practical
authority they have over gTLDs at this level of detail.    In
addition, while I don't know how something in this area would be
received, we have observed that some pronouncements from ICANN
have caused some relevant registries to tune out, regardless of
the merit of the advice/instructions given.

As one possibility, the WG might recommend that these script
groups publish best practices recommendations as RFCs, possibly
advising the IESG on ways to accept evidence of thorough study
and vetting within a community in lieu of expecting informed
consensus of the IETF community about the details of each script
and recommendation.

Perhaps there are other possibilities.  But, as the Arabic
Script IDN group (ASIWG) has discussed, there really are issues
that are best handled as registry policy (rather than in the
protocol) but that are more important than simple, everyday,
confusable character pairs.    If we can find an explicit way to
let the appropriate communities identify those cases and get the
word out -- a way that would permit referencing by bodies who
have authority to make policies as well as informing anyone
interested -- perhaps we can reduce pressures to incorporate
rules or actions into the protocol that do not logically belong
there.

        john