An idea for transition principles

[John C Klensin]

--On Monday, December 14, 2009 11:25 -0800 Gervase Markham
<gerv at mozilla.org> wrote:

> On 13/12/09 20:49, Vint Cerf wrote:
>> please see:
>> 
>> http://docs.google.com/Doc?docid=0AUIeBYngPBqkZGdxbnhzMmZfMTB
>> nNXg4OHRjcw&hl=en
>> 
>> Your comments will be much appreciated.
> 
> Vint,
> 
> I think this is excellent, but there is one point which might
> require  further clarification.
> 
> As Shawn has mentioned, client authors need to know how much
> time must  elapse before they can/should start switching their
> software to  implement IDNA2008 rules.

> Would the best approach here to be for IETF/ICANN to canvass
> the key  affected registries in order for us to determine the
> value of a time "S"  which represents the upper bound on how
> long they think it will take to  run the "sunrise" process?
> After that time, clients can begin to switch.  Other
> registries, while retaining full control over their own 
> operations, would know that if they did not run sunrise
> processes which  completed inside time S, resolution problems
> of various sorts may result.

Gerv,

We could ask ICANN to conduct such a poll.   I assume they would
be willing to do it, but Tina should comment on that --and how
rapidly they could do it and what sort of response she would
expect -- if she thinks that would be appropriate.

However, remember four things:

	(1) In theory, any TLD registry, anywhere in the world,
	can register labels containing German or Greek
	characters.  Those who are not doing so already can
	decide to start doing so at any time.  Some are
	obligated (or really strongly encouraged) to notify
	ICANN before doing so; others are not and, because
	"fuss" is not an IDN, it is hard to guess.   So the poll
	would need to be of all TLD registries and would need to
	include registries who have no transition problem (yet)
	because they don't register German or Greek labels
	(yet).  Note that a registry whose policies today don't
	permit registrations in German doesn't, in principle,
	need a transition strategy (sunrise or otherwise)
	because that registry cannot contain any label that was
	really intended to be Eszett or Final Sigma.
	
	(2) I gather that the timing of responses to such polls
	has a very long tail.  At some stage, one will simply
	need to decide when one has enough data to make an
	evaluation.
	
	(3) ICANN is in a reasonably good position to poll the
	TLD registries -- under 300 of them if I recall --
	although I can't guess at the expected value of the
	number of responses they would bet from that group.   If
	all of them respond (very unlikely, IMO), that is 300
	zones out of at least tens of millions and probably an
	order of magnitude more than that... a drop in the
	proverbial bucket if there ever was one.

	(4) There is no risk of any problem in a given registry
	as long as that registry doesn't accept actual
	registrations for these characters (i.e., register
	A-labels that contain them in encoded form rather than
	mapped to something else... or in the Eszett case, just
	accept ASCII labels containing "ss").  The high-risk
	period is between when such registrations start being
	accepted for a given zone and when the sunrise (or other
	procedure) has stabilized sufficiently that one can
	plausibly accept the risks.

Whatever those registries zones do, not all of those that care
about these characters (or may do so in the future) are going to
try to start introducing Eszett or Final Sigma at the same time
(whether they do it by sunrise or some other mechanism).  Some,
for example, may decide that they want to embark on some period
of customer education before doing anything.  Others may have to
consult with various policy bodies and could take months to even
decide.

There is also a bit of an interaction: the longer you folks wait
before making the transition, the more IDNA2003-style
registrations we will have that then need transition strategies
or pose risks.

So I suggest an alternate strategy: Decide.  Collaborate with
other browser vendors on an "earliest date" if you can, but
decide and announce your decision.   It almost certainly
shouldn't be before the IDNA2008 documents are published as
RFCs.  Lisa can probably give a better estimate, but I assume
that the earliest possible approval date is, at this stage, the
first half of January.  The earliest imaginable date for RFC
publication is 60 days after IESG approval.  That puts things at
least into mid-March and could easily be later.  Probably that
means that switching over for these characters should not occur
until at least 1 June or 1 July -- and maybe not until somewhat
later-- but that is ultimately _your_ business decision and that
of your colleagues/competitors.

Just my opinion; YMMD.

    john