Registry restrictions (was: Re: Domain names with leading digits (Re: Determining the basic approach))

[John C Klensin]

--On Monday, 05 May, 2008 18:50 +0200 Harald Tveit Alvestrand
<harald at alvestrand.no> wrote, responding to Paul Hoffman...

>...
>> Of course. But the WG is relying on registry rules and
>> recommendations  for many other things that are being changed
>> from IDNA2003. We need to  be consistent on this point: can
>> we remove things from the IDNA  protocol and have them be
>> enforced by registries, or can't we?

> Let's list them and do them case-by-case (in another thread,
> please).

Just to clarify the model that was used to shape the existing
documents, the assumptions are:

	* There will be things that cannot be done strictly in
	the protocol that are, nonetheless, important.  We hope
	that they can be handled by registry restrictions and
	believe that registry restrictions (heavyweight or
	lightweight and independent of the methods used) are an
	important part of the picture, not just desirable.
	
	* There will be things that cannot be done strictly by
	registries that are, nonetheless, important.  We hope
	that they can be handled by protocol provisions (at
	registration time, lookup time, or both) and believe
	that such protocol provisions are an important part of
	the picture, not just desirable.
	
	* So, even though we are not, in general, prepared to
	tell registries what provisions and restrictions they
	should (or must) apply, we believe that some things are
	better done by registries than by protocol.
	
	* We also believe that there are some things that
	dangerous or treacherous enough, or that raise long-term
	issues that are potentially problematic enough, that
	they have to be checked in the protocol, not just left
	to registration-time activity.  The most important
	example of this involves characters that are invisible
	out of context, notable the joiners, but we believe that
	special measures for unassigned and disallowed
	characters are also in order.

That obviously leaves three sets of questions for the WG:

(1) Is that model reasonable?

(2) What very specific things should be pushed off to the
registries (remembering that "registry" equals "any zone in the
DNS") and which ones left in the protocol (at registration time,
lookup time, or both)?

(3) For any given thing we try to push off to the registry, are
we offering advice or do we think we have some leverage on
telling registries what they "MUST" do?

Just IMO.

     john