minutes from the newdom BOF

["Rick H. Wesson" <wessorh@ar.com>]

Attached are the minutes from the newdom BOF held at the last IETF meeting.
I am sending them in the hope that thy refresh everyones memory of
what was said there. Many thanks to Stan Barber <sob@academ.com>
whom took the minutes.

-Rick


-------------------------- cut here ----------------------------------

These are my notes from this BOF. They have not been edited in any way.


New Top Level Domain BOF
37th IETF

Rick Wesson <rick@ar.com>

Overview
Discussion
Shared Domain Registries
draft-higgs-cat-03
RIDE -- Registry Information Database Exchange
Discussion
Viability/Charter for Working Group

Rules --
	Respect each other
	Moderated Discussion
	Raise your hand if you want to speak
	Moderator will allocate time to topics
	What you get out of this is directly related to what you contribute

Discussion List
	Governance is a shared registry
	Should all iTLDs be shared?
	Registry performance
	Can a registry grant ownership of the namespace?
	Can a registry charge fees?
	Should there be a limit on fees charged by registry?
	Should there be new TLDs?
	Should iTLDs be shared? Is there a precedent for non-shared?
	What about existing applications? Are these applications for TLDs of
	registry?
	What is the problem we are trying to solve?
	In that context, what are the future directions?


For background, check http://www.iahc.org.


Shared Registries

A Shared Domain Registry

CORBA
Java Management Interface
Database Abstraction
12,000+ lines of C/C++ Multi-threaded Server
Granularity of Object Locking
Client Template Parsers (InterNIC and RIPE)

A series of domain allocators are able to use this common registry without
stomping on each other.

OMG CORBA 2.0 was used to develop this.

The core technology that makes it work is the Common Object Request Broker

Why not use DNSUPDATE? There is alot of information used in the registry
that is not captured in DNS.

IIOP -- Internet Inter Orb Protocol

IDL is used to define the interface for this environment. This can be used
to generate code in JAVA, C/C++, etc.

This makes three tier system: Client/Object Servers/Database

Why not CGI? There are a large number objects at the InterNIC and using
CGI is hard. Real databases and really well managed systems have to support
such large databases.

NEO/JOE was used. NEO is used to build IDL. JOE is used to translate this into
JAVA.

There are free tools that can do this as well. In fact, JOE is freely
available.

Goals:
Database abstraction -- Any SQL or OO database could be used.
Object and Attributes -- It is independent of language. The composition
and attributes of the objects can be described.
Authentication -- Not there yet. Next Release in January.
Locking -- Multi-Threaded Server with attribute level locking.
Parsers -- written in Lex for domain, host, contact and ACK/NAK templates and
one in Java that actually could create the records live.

This software does not really change the current situation.

How does this handle policy? It doesn't handle policy. Mike O'Dell argues that
is really a way to support back-office operation. It is like the mechanism that
currently supports 800 number allocation.

How is the cost of operations shared? Mike notes that in an RFP response
several years ago, there was a suggestion that folks would pay for what they
wanted and those fees would be allocated such that the central registry
would get their support from these fees.

This is not a replicated databases, it is a central database with atomic
highly granular locking.

http://sentosa.ar.com

Simon Higgs speaks about his draft.

The central proposal is to create three TLD classes: Shared (any registry can
participate), non-shared (one registry), private (one organization)
Simon asserts that MIL and GOV are private under this systems.

Expansion of the name space

1. Existing TLDs placed into classes
2. New TLDs created by IAHC
3. Trademark issues addressed by International Schedule of Goods and Services

Each TLD should have a charter
	RFC 1591
	Domain Dispute Policy
	Voting share arrangement for domain holders

Rights to names
	Dispute policy
	Whatis generic

Registry Selection
	Operational Guidelines by Class

Comments on Trademarks -- Each Country has specific issues. There is also an
issue of which organization has jurisdiction.

There are always going to be local jurisdiction.

Paul M. suggests that the use of public keys might mitigate this.

The voting mechanism is a possible can of worms, but the mechanism is not yet
developed. It has been suggested that giving the voting group the ability
to change the charter is not a good idea. There is also

Is there any objection to the classes? Who decides? Should not the charter
determine that? Who writes the charter? What prevents a commercial organization
from being TLD?

Another person suggested that the real key to this is the development of
the charter. The classification is not useful. Simon argues that he has
found any current domains that cannot be described as being in one of these
classes.

Now comes David Kessens <davidk@isi.edu>

Registry Information Database Exchange formats (RIDE)

It is very difficult to get all the data from everyone who does registration.

What will RIDE do?

1. Define standardized dataformats for Internet registries (IP, routing, domain
registries)

2. Define standardized dataformats for contact information of those
registries

3. Define how one can access and find this data

4. Eventually, define a distributed system to access and also modify this data

Subscribe to "ride@isi.edu" via "ride-request@isi.edu", for the possibility
of creating a working group to define this format.

Should there be new iTLDs?

	There is generally sentiment that there is demand for them. There is
some
concern that having them will create problems for existing country-based
top-level domains.



Should ITLDs be shared?
	Yes, some ought to be. However, there is a large liability issue for
	everyone involved. David Conrad argues that only the entity that put
the
data in incorrectly
	would be liable.
	Mike wants to know the value of a non-shared TLD – The EGB answer is
that
someone will
	be making money. Mike want to know the technical or operational
advantages.
Simon believes
	that the charters may not be enforceable. Mike says there are two
levels of
sharing. One level is
	that where the data is exclusively controlled by one organization and
multiple registrars. Another
	is the sharing of the data itself. The former model would permit a
point of
control on admission.

	Exclusivity promotes commercial viability.  Exclusivity is good for
small.
It is good for a vertical
	markets.

	Christian comments that everything could go as a second level domain.

	What happens if a registry screws up? It is argued that should occur,
that
registry should loose
	its "license" to do registration.

	This seems to be like telephone number portability.

	Simon argues that all TLDs should be shared, even the country code
registries. Scott Bradner
	argues that it not the case for some countries.

	Someone from Dunn & Bradstreet offers that they maintain a registry of
many
records. They
	have had to become a legal entity in 166 countries. They use a
directory to
match the entities
	to their registration. He offers that TLDs that are only serving a
small
number of users is really
	bogus.

What about existing applications?  Are these applications for TLDs or
registries?

	Eugene says that they should be thrown out. There is also the possible
use
of the concept
	of  "prior use"….

	Others suggest that they had expectations that these were applications
to
operate registries for
	top-level domains.

	Simon says he has another draft that will address this issue to some
extent.

Can a registry grant ownership of the name space?

	It is an irrelevant question because you can’t define ownership. Is
ownership the correct word?
	Probably not. Who has the authority to change the delegation? Mike
suggest
that the ownership
	is really stewardship.

	Bill Manning puts for the notion that the operation of the registry is
for
the public good.

	What about the notion of the allocation of a domain name without
putting it
into the DNS?


Can a registry charge fees?

	Yes. In Germany, there is a large fee for registration. There is much
interest in getting these costs
	down.

Governance of a shared registry

	How would one set up an organization that would prevent a monopoly from
occurring?

Registry Performance

	The registries should be interoperable and robust. Scott Bradner offers
that the current work on
WYPO? Http://ksg.www.harvard.edu/iip/intellec.html will complicate this
work. Paul M. notes that the State Department argues that you need to have
prior authorization to cache information (like DNS entries). This is not the
position of many other parts of the U.S. Government, but it does present

Mark Kosters announces that tomorrow there is an rwhois working group
meeting in which some of these issues will be discussed.

Has there been any thought given to measuring a registries performance?

Problem Definition

	What is the problem?

	One person suggests that more registries would help distribute the
load.
Another person comments that of the 180 registries, they themselves believe
that the InterNIC does better than any of them.

	Another person argues: "The problem is really a directory problem.
Expanding the domain space is only going to make it worse." Essentially, DNS
is being used as a directory service. Isn’t ½ the InterNIC supposed to deal
with that? AT&T has been doing some work in this area.

Paul Dixie agrees that DNS is being overloaded, but supports the idea of
having more iTLDs since doing that will push the IETF to create a real
directory service.

Scott Bradner also sees the need for a directory service. He is organizing a
workshop to look into this. Paul M. suggests that the web page search engine
guys have taken over . Another guy suggests that
the advantage of directory services is precision. There is also "business
card" issue. X.400 failed because of this. Mike suggests that information
retrieval technology exists that could help the web folks do things much
better. Another person suggest that this is really a library science
problem. Another suggest that the problem is really a quality of information
problem.

Someone from Intel suggests that a mechanism for directories should be
developed to foster information sharing.

Other suggested that as the DNS continues to be used in the fashion, it will
mandate the directory solution be created.

Should there be a working group for the non-technical work?

It appears not.





-- 
Rick H. Wesson