Another round of IDNAv2, and thoughts on IDNA2008 goals

Simon Josefsson simon at josefsson.org
Wed Mar 4 23:39:31 CET 2009


Paul Hoffman <phoffman at imc.org> writes:

>>One question: do you think there are input strings that are not allowed
>>by IDNAv1 but will be allowed by your IDNAv2 that will be encoded
>>differently in IDNAv2 compared to IDNA2008?  I am aware that ezset may
>>be one example, but I'm looking for other examples.
>
> Your phrase "are not allowed by IDNAv1" needs to be better
> defined. Esszet *is* allowed by IDNAv1: it is mapped to "ss". In the
> current drafts of IDNA2008, four characters have that property in
> common: Esszet, Greek final sigma, ZWJ, and ZWNJ.

Right, I meant to ask two question there, but if IDNAv2 will never
co-exists with IDNA2008 these questions are irrelevant.

>>If there are many examples of these strings, rather than the few
>>exceptions like ezset, I think we are better of discussing whether to
>>completely abandon either IDNAv2 or IDNA2008, since making
>>IDNAv1+IDNAv2+IDNA2008 work together will be a complex task.
>
> I do not want a world with IDNAv1+IDNAv2+IDNA2008; it should either be
> IDNAv1+IDNAv2 or IDNAv1+IDNA2008. I deeply apologize for not making
> that clearer.

I had not understood that.  Then my opinion of IDNAv2 right now is less
positive, and here is why.  Everyone agrees that there are many problems
to solve in IDNA2003.  If IDNAv2 is the end of the road in addressing
these questions, it would be like giving up on the harder questions that
IDNAv2 dodges for.

It is not clear to me that we have reached a point where we cannot
resolve all problems.  Or even a point where we have understood what we
are trading the current problems for; IDNA2008 may introduce other
problems.  It seems IDNA2008 is relatively close to being implementable
and tested in practice.  If we can move a bit further on that road,
we'll find out whether we have a good trade-off in IDNA2008.  If we find
that we cannot solve some issues without introducing other equally
difficult problems, then going back to IDNAv2 would be a wise path.

Thus I'm leaning towards that it may be too early to consider IDNAv2
now, but I applaud you for making us think about this now rather than
later in the IETF process when considering options will be more
difficult.

I think your point about re-chartering the WG is valid.  I haven't been
participating here a lot, but I was involved in the chartering
discussion, and I recall that my perception was that the WG will attempt
hard to be backwards compatible or re-charter.  As far as I can tell
IDNA2008 is not fully backwards compatible so the WG should re-charter.

I think IDNAv2 should continue breath down the neck of IDNA2008 so that
we have an option if IDNA2008 doesn't complete and result in a good
trade-off.  So I also believe it is too early to say no to your work.

/Simon


More information about the Idna-update mailing list