[wg-c] position paper

[Jonathan Weinberg <weinberg@mail.msen.com>]

	This is an initial draft of a position paper expressing a few of my own
views and reaching towards consensus positions.  Notwithstanding that I've
pulled it together in an attempt to enunciate positions with wide appeal,
I'm offering it wearing my "working group member" hat, not my "co-chair"
hat.  I'm happy to fuzz out some issues in order to reach agreement with
others. To the extent that any of you would be comfortable signing one part
of the document but not another, let me know, and I'll try to figure out a
way to accommodate that.  Various WG members have already seen a version of
this document.

	The question of new gTLDs involves about half-a-dozen closely
inter-related issues.  I try to work through those issues one at a time.
The position paper doesn't present specific implementation plans for every
aspect of gTLD expansion.  We've been sharply divided on issues of basic
principle, and in this position paper I'm primarily trying to work towards
some settlement on those broad-brush issues.  Once we agree (or agree to
disagree) on those, we can move on to specific implementation details.

	This position paper, in summary: ICANN should add a large number of new
gTLDs, including both general-purpose and limited-purpose domains.  It
should begin with an initial testbed rollout of six to ten new gTLDs, with
the expectation that it will add more if the evaluation period reveals no
major problems.  In the short term (during the testbed and initial phased
rollout), to the extent that more registries seek immediate inclusion of
their preferred gTLDs into the root than the schedule will accommodate,
ICANN may engage in an ad hoc selection process.  In the long term, what is
most important is that ICANN adopt policies minimizing the barriers to
adding new gTLDs to the root.  ICANN should authorize both for- profit and
non-profit gTLD registries.  It should enunciate a rebuttable presumption
that all gTLDs (whether for-profit or non-profit) must support competitive
registrars.

I.	Should there be new gTLDs?

	Yes.  Expanding the number of TLDs will increase consumer choice, and
create opportunities for entities that have been shut out under the current
name structure.  It doesn't make sense to continue a situation in which
huge segments of the general public view .com as the only "real" TLD, but
every word in a typical English- language dictionary is already registered
as an SLD there.  This situation is currently requiring companies to
register increasingly unwieldy domain names for themselves, and is
inflating the value of the secondary (speculators') market in .com domain
names.

	Right now, .com stands astride the name space as the dominant -- indeed,
the only "real" -- commercial TLD.  Companies that currently have a domain
name in the form of <www.companyname.com> have an extremely important
marketing and name-recognition tool.  They have an advantage over all other
companies that do not have addresses in that form, because they are the
ones that consumers, surfing the Net, will be able to find most easily.  If
the name space is expanded, companies will be able to get easy-to-remember
domain names more easily, and the entry barriers to successful
participation in electronic commerce will be lowered.  Similarly, addition
of new gTLDs could enlarge noncommercial name space.

	Addition of new gTLDs will allow different companies to have the same
second-level domain name in different TLDs.  That is (to pick an arbitrary
example), shopping.com might face competition from shopping.biz and
shopping.store.  Those businesses will have to compete based on price,
quality and service, rather than on the happenstance of which company
locked up the most desirable domain name first.

	Indeed, expanding the name space could help solve one of our most
intractable problems relating to trademark and domain names.  Currently,
when multiple unrelated companies have the same or similar names (such as
United Airlines and United Van Lines), there is no good way to resolve the
question of who gets the valuable domain name <www.companyname.com>.  But
if the domain name space were expanded, so that one firm could have, say,
<www.companyname.biz> and another could have <www.companyname.firm>, many
of these disputes could be avoided.  Consumers, understanding that a given
SLD string can belong to different companies in different TLDs, would be
less likely to jump to the conclusion that any given domain name was
associated with a given company.

II.	What should be the nature of the new gTLDs?

	Some people argue that new gTLDs should be predominantly or entirely
limited-purpose domains (that is, .airline rather than .firm).  Their
arguments are that limited-purpose TLDs give more information to the
consumer (a consumer can expect that united.airlines is an airline, not a
trucking company) and will be less threatening to trademark interests
opposed to the expansion of the name space (because United Airlines will
not feel threatened by the registration, say, of united.books).  Some even
argue that ICANN should map out, at the start, a framework of
limited-purpose TLDs, such as .transp and .health, so that users could rely
on the structure of the DNS in seeking the URLs associated with particular
businesses or content providers.

	I agree that limited-purpose TLDs can be useful.  It seems to me, though,
that these should not be the *only* new gTLDs.  For the reasons I stated
above,  .com is currently the 500- pound gorilla of TLDs, and domain names
in .com have tremendous (artificial) market value.  Adding a set of
limited-purpose TLDs would not change that: there would remain .com (and to
a lesser extent .net and .org) atop the TLD pyramid, and a mass of
special-purpose TLDs below.  Alternative *general-purpose* top-level
domains, by contrast, could provide effective competition to .com.  This
would more nearly level the playing field for individuals and businesses
seeking attractive domain names, and would diminish the ability of a
minority of e- businesses to collect rents based simply on their
registration of good names in the "best" TLD.

	This suggests that the ideal system would be one that mixes new
limited-purpose gTLDs with new general-purpose ones.  Part VII of this
position paper discusses the process through which new gTLDs will be
selected.  If ICANN adopts a process in which it selects the gTLD strings
itself, it could simply choose some gTLDs in each category.  Alternatively,
if ICANN adopts a process in which it simply accredits new registries and
those registries choose their own TLD strings, the result would likely be a
similar mixed system.  Some registries might seek broad market reach by
choosing a name with general appeal, such as .web; others might seek to
exploit a niche market by choosing a name with limited uses, such as .per.
The resulting name space would have gTLDs in each category.

III.	How many new gTLDs should there be?

	In the long term, a domain name system with a large number of gTLDs seems
highly desirable.  Each additional TLD provides new options for domain-name
registrants.  Further, a name space in which there are many TLDs and TLD
registries is one in which the registries themselves are (or can be)
subject to useful market competition.  The existence of many competing TLD
registries will diminish the market power that any particular TLD will
exercise.  Users who are unimpressed with the performance of one registry
can instead acquire a new domain name in a different top-level domain, run
by a different registry.  Members of the Internet community will be able to
decide themselves, through their name-registration decisions, which of the
many possible TLDs will play an important role in the name space.

	Imagine, thus, that the name space contains a single registry called
.sports.  Without more, this registry has market power corresponding to the
degree that it is a better TLD for certain registrants than any other, and
it can use that market power to extract inefficient rents.  If, on the
other hand, there are many TLDs, and relatively free entry into the TLD
namespace, then .sports can constrained by the creation of .athletics, to
the benefit of domain-name registrants.    In this respect, it would be
undesirable to artificially limit the number of gTLDs short of the bounds
of the technically feasible and operationally stable.

	The most important reason that has been suggested for limiting the number
of gTLDs relates to trademark owners' concerns.  Addition of many new
gTLDs, trademark owners urge, will greatly increase their policing costs.
The more new gTLDs there are, they argue, the more work they will have to
do in order to ensure that nobody is using their trademarks (or variants)
as second-level domains in ways that would confuse consumers.

	This concern, it seems to me, is overblown.  Trademark owners are already
policing their marks in the existing generic top-level domains, as well as
in a variety of country-code top-level domains. Cost concerns can best be
addressed through requirements that the new top-level domain registries
make their lists of second-level domains easily searchable through an
automated process.  Such a rule will limit the costs to trademark owners of
adding new domains.  More importantly, it would not make sense to distort
the entire structure of the Internet name space simply in order to avoid
additional costs to trademark owners.

IV.	What should the transition to an expanded namespace look like?

	It would not be sensible to dump a huge number of new TLDs into the
namespace on Day One.  Rather, ICANN should supervise a smooth transition
to an expanded namespace.  During that transition, ICANN should require any
would-be registry to satisfy qualifications relating to technical
competence and stability.  Further, it should limit each qualified new
registry, at the outset, to a single gTLD.  This will allow the greatest
number of entities to participate in the transition, and generate the most
"bang for the buck" in new registry competition.

	ICANN can choose either of two transition plans.  First, it could announce
a testbed in which a fixed number of new registries would be added to the
root, followed by an evaluation period.  Assuming that the testbed revealed
no major problems, it would then move to a phased rollout in which new TLDs
would be added to the root as qualified registries applied to run them, but
no more than a fixed number of registries would be added per month.  The
per-month "cap" would ensure that new TLDs would not be added so quickly as
to overwhelm ICANN's processes or the new dispute- resolution system.

	Alternatively, ICANN could dispense with the testbed, and simply begin
with the phased rollout.  Under either plan, after the close of the
transition period, ICANN could establish procedures not only to add new
registries, but to allow existing registries to add additional TLDs.

	During the Working Group C straw poll, a large number of group members
urged that dispensing with the testbed would be imprudent.  In discussions
building on that straw poll, the dominant (though not unanimous) sense of
the working group was that even though we had differing views as to the
*best* approach, almost all of us could support a compromise proposal that
ICANN begin with an initial testbed rollout of six to ten gTLDs, with the
expectation that it would add more if the evaluation period reveals no
major problems.  I support that rough consensus as a workable resolution of
this question.

V.	Should ICANN require each new gTLD registry to be shared — that is, to
support competing registrars on an "equal access" basis?

	In a shared-registry system, the registry must support multiple
registrars, subject to requirements designed to ensure that the registry
does not exclude qualified registrars or give some registrars
higher-quality access than others.  This approach assumes (as ICANN
currently does in connection with the NSI gTLDs) that the functions
associated with SLD registration can be divided into two: the "registrar"
function of dealing with the end user, and the "registry" function of
maintaining the registry database.  Sharing ensures competition, on a level
playing field, in the provision of registrar services in any given TLD, and
gives the user of the benefits of competition with respect to those
services.  In particular, by ensuring that somebody who is skillful at
providing registrar services will be operating in each TLD, it provides
some protection against the situation in which the registry controlling a
desirable TLD is clueless at providing registrar services.  These are
significant advantages.  The Nominet system in .uk is an example of a
highly successful shared-registry system

	On the other hand, sharing is not a panacea; in particular, it has no
effect on the registry's market power as a registry.  (That is, the
requirement that a registry support multiple registrars does not itself
constrain the price that the registry is able to charge for registry
services.  That price can only be limited by market forces or -- as in the
case of NSI -- by direct regulation.)  Further, it is possible to imagine
situations in which a sharing requirement would actually interfere with the
provision of innovative services.  Consider, for example, a plan to operate
.family, limiting registration to persons operating Web sites that in the
considered (discretionary) judgment of the TLD operator were
family-friendly.  The commercial success of that domain would depend in
part on the public's view of the care and discernment the operator
exhibited in making those choices.  That business plan would not succeed if
any registrar could add names to the .family domain in the exercise of its
own discretion.

	In light of these concerns, it would be sensible for ICANN to enunciate a
*presumption* that all TLDs support competitive registrars, thus generating
the benefits of registrar competition, rebuttable to the extent that a
particular registry makes a showing to ICANN there would be significant
advantages to some other arrangement.

VI.	Should ICANN require that each new gTLD registry be operated on a
non-profit (cost- recovery) basis?

	Both non-profit and for-profit registries have advantages.  The advantage
of non-profit registries lies primarily in the expectation that they will
be less inclined to exploit their market power at the expense of
domain-name registrants.  The advantages of for-profit registries lies
primarily in their greater incentives to operate efficiently and offer
innovative services.

	Registries have the potential to wield significant market power, for two
reasons.  First, any registry will have market power corresponding to the
extent to which the top-level domain it controls is more attractive, to
some subset of users, than are the alternatives.  Put another way, each
registry has "monopoly" control over its particular domain(s).  NSI, thus,
would be able to extract monopoly rents flowing from its control over .com,
were the price it charges for .com registry services not directly regulated
by the DOC-NSI cooperative agreement.  Second, a registry will have market
power over its existing users by virtue of "lock-in": that is, to the
extent that it is difficult for a user to switch to a new registry (because
it would mean abandoning an existing domain name), a registry will be able
to exploit users' reluctance to move.

	The primary advantage of non-profit registries is that non-profits may be
less inclined to exploit their market power by charging monopoly rents.
This advantage is not guaranteed. Non- profits often do not have great
incentives to reduce costs and prices; a non-profit registry, not subject
to rigorous cost control, might simply allow its costs to balloon and raise
prices as necessary to cover them.  At least some non-profits, though, may
do an excellent job of providing registry services at low cost.  Again,
Nominet is exemplary.

	For-profit registries have a different set of advantages.  Most
importantly, the registry's interest in making more money will lead it to
be more energetic in seeking to heighten efficiency, lower prices and
provide additional value-added services.  Similarly, a for-profit registry
will likely be much more energetic in branding and marketing the TLD.  As a
general matter, one can expect for-profit TLDs to be more likely to come up
with innovative services and innovative structures, to the benefit of users.

	Other mechanisms may address the problem of market power. Remember that a
registry can derive market power from two possible sources.  To the extent
that that market power simply derives from the attractiveness of the TLD,
the expansion of the name space will itself provide an answer: As the
number of gTLDs expands, the attractiveness of any particular gTLD
vis-a-vis alternative, competing gTLDs will diminish.  In the current
architecture, in which we have only a few gTLDs, users have few credible
alternatives in choosing a top-level domain.  To the extent that users have
many alternatives, on the other hand, TLD registries will wield little
market power, and the premium that a TLD registry will be able to charge in
return for registrations in that domain will shrink.  For-profit firms in
competitive markets have substantial incentive to reduce their costs and
prices; specifically, in a competitive environment, a registry that wants
to attract new registrants will keep its costs and its prices low.

	To the extent that a registry's market power derives from lock-in, the
expansion of the name space will not provide an answer.  There may,
however, be other constraints.  At least so long as the market is growing,
a registry will likely have little incentive to put the screws on existing
customers, because that would discourage new customers.  Other mechanisms —
possibly including the availability of long-term contracts and registry
codes of conduct discouraging sharp price increases, as well as mechanisms
making it easier for users to change domain names — may also play a useful
role in addressing lock-in.

	In light of these factors, it would make sense to have a mixed system of
nonprofit and for-profit registries.  Members of each group would have
something different to offer.  Domain name holders most concerned about
lock-in would be free to use domains run by nonprofits if they thought that
a safer course; other folks could take advantage of the dynamism and
innovation of the for-profits.

VII.	What should ICANN's process be for selecting new domains and registries?

	It is useful to address this question from both a long-term and a
short-term vantage point.  In the long term, ICANN can take one of two
approaches.  Under the first approach, ICANN would select new gTLDs to be
added, and then solicit applications from would-be registries to run those
gTLDs.  Under the second, ICANN would accredit registries by reference to
criteria including the applicants' technical abilities to perform registry
services and  financial stability, and the registries would then decide for
themselves what the names of their gTLDs would be (subject to a process
under which ICANN could resolve conflicts, and could deem certain gTLD
strings out of bounds).

	Notwithstanding that these two approaches are often presented as radically
different, they are more similar than they appear.  Under either approach,
a relatively small group of people will choose the names of the new gTLDs.
Under the first, the decision will be made by ICANN decision-makers, on the
basis of their views as to which new gTLDs would be most beneficial to the
community.  Under the second, the decision will be made by registry
operators, on the basis of their views as to which new gTLDs are most
desired by the community (that being the course that will generate the most
registration dollars).  These are similar considerations, and will likely
result in similar sets of names.

	Under either approach, if ICANN lets *enough* new TLDs into the root, it
will be the Internet users themselves who ultimately decide which gTLDs
succeed and which will fail.  Users will make their own choices as to which
top-level domains to register names in; as a result, some new gTLDs will
thrive, and some will stagnate.  It will be the community that decides.

	In that respect, ICANN's choice between these two approaches is not
crucial.  What is most important is that ICANN adopt a policy under which
it is relatively easy to get new gTLDs into the root (whether operated by a
newly created registry or an already-existing one).  Thus, ICANN's criteria
for the qualification of new registries should not be unreasonably
burdensome.  Moreover, to the extent ICANN passes directly on the content
of a proposed new TLD string, its processes should be geared towards
approval without extensive delays.  Finally, aside from its reasonable
limitation on the *pace* of gTLD expansion during the phased rollout
period, ICANN should not seek to limit the total number of gTLDs short of
the bounds of the technically feasible and operationally stable.  If
ICANN's process satisfies these criteria, then either of the selection
processes described above will work.

	The remaining important question is that of the short-term: How should
ICANN proceed during the testbed and phased rollout, to the extent that
more registries seek immediate inclusion of their preferred gTLDs than the
schedule will accommodate?  There is no especially satisfactory answer to
this question; by hypothesis, some qualified applications will have to be
deferred.  None of the available paths to decision are ideal.  The process
that would likely make ICANN itself most comfortable would be an ad hoc
selection process under which ICANN was free to look to a variety of
factors, including both the qualifications of the registry and the nature
of the TLD.  In the long term, such an approach would not be desirable; it
presents the risk of subjective and unaccountable decision-making.  In the
short term, however, as a matter of supervising the initial rollout, it
would be responsive to oft-expressed concerns about Internet stability and
reliability.


Jon


Jonathan Weinberg
weinberg@msen.com