Comments of Jonathan Weinberg
Professor of Law, Wayne State University
weinberg@msen.com
I
am submitting these comments in my personal capacity. Although I was the co-chair
of the DNSO's Working Group C on new generic top-level domains, these comments should
*not* be taken as representing the views of the working group, or of anyone other
than myself.
Q1: In the introduction of new TLDs, what steps should be taken
to coordinate with the Internet Engineering Task Force, the Internet Architecture
Board, and other organizations dealing with Internet protocols and standards?
No
such steps need be taken. From the perspective of Internet protocols and standards,
the introduction of a new gTLD is no different from the introduction of a new ccTLD.
Q2:
What stability concerns are associated with the initial phases of registration within
the TLD?
The word "stability" tends to mean different things when used by
different people. For me, "stability" refers to whether the DNS will work —
whether domain names will resolve to IP addresses as they should. The introduction
of new gTLDs should have no effect on IP address- domain name resolution for names
in the old TLDs. On the other hand, if the software or hardware for a new TLD
registry is bad or underpowered, that could cause names in the new TLD to fail to
resolve. If a new TLD's shared registry system malfunctions, that could (in
a worst case analysis) lead to conflicting registrations in that TLD. The possibility
of large query and transaction loads when a new gTLD opens up increases the risk
that a new registry will be unequal to its task.
Q3: What can be done to eliminate
or reduce these stability concerns?
New registries can be subject to technical
requirements relating to physical plant, scalability and load capacity, and data
security and escrow, and can be required to show proofs of financial responsibility.
Q4:
Would these stability concerns be magnified by introducing a large number of TLDs
at once?
If many new registries go on-line simultaneously, the possibility that
one will suffer technical difficulties may be increased. The technical requirements
I mentioned in my answer to Q4, though, should adequately address that concern.
Q5:
Are there any practical means of reversing the introduction of a significant new
TLD once it goes into operation?
No.
Q6: Is it feasible to introduce a
TLD on a "trial basis," giving clear notice that the TLD might be discontinued after
the trial is completed?
No.
Q7: To ensure continued stability, what characteristics
should be sought in a proposed TLD and in the organization(s) proposing to sponsor
and/or operate it?
The characteristics of a new TLD should not affect stability
(as defined in my answer to Q2), except that a TLD attracting a high volume of registrations
may be more likely to have scalability, etc. problems associated with its rollout.
It would be seriously misplaced, though, to avoid deployment of a popular TLD on
this basis. The technical requirements mentioned above can adequately address
these issues.
If ICANN selects an organization to "sponsor" a TLD, that organization
should itself be financially responsible and stable.
* * *
Q10: What lessons,
if any, can be learned regarding new gTLD introductions from the experience of the
ccTLD registries?
From a technical standpoint, new ccTLDs are indistinguishable
from new gTLDs. Most importantly, the ccTLD experience teaches us that the
introduction of new TLDs has no effect on the technical stability of the larger system.
The ccTLD experience also teaches that the mere introduction of a new TLD does not
guarantee commercial success or high registration volume. Although several
ccTLD registries have sought to market themselves as competitors to .com, for the
most part they have not succeeded, because they have not been able adequately to
capture worldwide public awareness. I believe that competitors to .com can
and will succeed, but any TLD seeking to accomplish that goal needs to recognize
the marketing and branding challenge it will face.
Q11: Can lessons relevant to
introduction of new TLDs be learned from the recent decisions by a number of them
to operate in a globally open manner? If so, what lessons?
See above.
Q12:
Is the Names Council's recommendation that a "limited number of new top-level domains
be introduced initially" a sensible way to minimize risks to Internet stability?
It all depends (with apologies to President Clinton) on what the word "limited"
means. It certainly makes sense not to dump several hundred new TLDs into the
root all at once. New TLDs should be introduced in such a manner that ICANN
can learn from the initial introduction. At the same time, the initial rollout
should not be too small. A too-small rollout will not yield useful information.
It will exacerbate the land rush and name speculation that some fear will accompany
deployment of new TLDs, since registrants' attention will be entirely concentrated
on a too small, too constrained set of new TLDs.
Q13: What steps should
be taken to evaluate carefully the initial introduction of TLDs before future introduction
of additional TLDs?
ICANN should simply consider whether the initial introduction
of new gTLDs has created serious problems. If not, it should proceed with further
deployment. If so, it should consider how those problems can be addressed in
later rollouts.
Q14: Should a fixed time be established for all the evaluations,
or should the time allowed vary depending on the nature of the TLD and other circumstances?
What
is most important is that a fixed timetable for evaluation (and hence for additional
deployment, assuming that the results of the evaluation are positive) be established
in advance. It will have taken five years after the initial draft-postel to
deploy the first new gTLDs. A fixed schedule will help ensure that the next
set too is not interminably delayed.
Q15: Should choices regarding the types
of TLDs included in the initial introduction seek to promote effective evaluation
of:
* the feasibilty and utility of different types of new TLDs?
* the efficacy
of different procedures for launching new TLDs?
* different policies under which
the TLDs can be administered in the longer term?
* different operational models
for the registry and registrar functions?
* different institutional structures
for the formulation of registration and operation policies within the TLD?
*
other factors?
It would be useful for the initial introduction to include TLDs
that allow the effective evaluation of each of the factors mentioned in this question.
Q16: Should any particular goal for, or limit on, the number of TLDs to be included
in the initial introduction be established in advance, or alternatively should the
number included in the initial introduction be guided by the extent to which proposals
establish sound proofs of concept of varied new TLD attributes?
Ideally, ICANN
should establish relatively hard-edged criteria against which to evaluate proposals,
and should announce that all proposals securing a passing grade when measured against
those criteria will be approved. A second-best approach, if ICANN believes
that it must limit the number of TLDs in the initial rollout, would be to announce
the target number of new TLDs at the outset. The least desirable approach would
be to announce vague criteria, without a target number, so that the evaluators would
have ad hoc discretion both as to the number of new registries and their identity.
Q17: In view of the current competitive conditions, should the promotion of effective
competition in the provision of registration services continue to be a significant
motivation for adding fully open TLDs?
Yes.
Q18: Should the desire for
diverse vendors of registry services in open TLDs be an important motivation in adding
fully open TLDs?
Yes.
Q19: Would the introduction of additional undifferentiated
TLDs result in increased inter-TLD confusion among Internet users?
Except
in the very short term, it should not. Today, especially in the United States,
.com is by far the dominant TLD. This dominance has made inter-TLD confusion
more of a problem: Many consumers, upon hearing any domain name, assume that
it *must* end in .com. Successful introduction of new TLDs, however, will obviate
this problem. The consumer will quickly realize that a URL may end in any of
a variety of domain names, and will understand that example.newTLD and example.com
refer to different resources.
Q20: Taking all the relevant factors into account,
should one or more fully open TLDs be included in the initial introduction?
Yes.
Q21:
How many?
The initial rollout should include at least three fully open TLDs;
more would be better. There are varying considerations here. On the one
hand, a single new TLD would be best positioned to provide competition to .com.
On the other hand, by virtue of that very fact, attractive SLDs in such a single
new TLD would be quite valuable (more than the likely registration fee), and the
TLD would be mobbed by speculators. By contrast, if ICANN authorized a large
number of open TLDs, the law of supply and demand illustrates that the value of attractive
domains in those TLDs would be lower, and speculation would be less of a problem.
Q22:
How effective would other fully open TLDs be in providing effective competition to
.com?
This will depend on the marketing, branding and business plans of
the registry proprietors.
* * *
Q25: Is increasing the utility of the
DNS as a resource-location tool an appropriate goal in the introduction of new TLDs?
The
DNS does a bad job as a resource-location tool, and always will. It does a
pretty good job as a *mnemonic* tool. That is, it is better suited to helping
a user remember the location of a particular resource, than to helping the user find
that resource in the first place. It is appropriate, in introducing new TLDs,
to ask whether a particular TLD will enhance the utility of the DNS as a mnemonic
tool.
Q26: Would the introduction of unrestricted, undifferentiated TLDs
run counter to this goal?
The introduction of unrestricted TLDs will not
significantly enhance the utility of the DNS as a resource-location tool or a mnemonic
tool. It will serve other purposes: increasing the supply of attractive domain
names (see infra), increasing competition in the domain name space, and combatting
the unhealthy domination by .com of the name space. Because the DNS will never
be a great resource-location tool, the fact that unrestricted TLDs will not significantly
advance that goal should not be seen as especially problematic. It is a hopeless
task to try and make the name space too neat and orderly; the real world is not so
neat, and the name space should mirror the real world.
* * *
Q28: Is the concept
of TLD "charters" helpful in promoting the appropriate evolution of the DNS?
In a world in which there were many TLDs, and in which it was easy for a new
proposed TLD to gain access to the root, it would be desirable for a TLD proprietor,
if it desired, to draw up a "charter" explaining the nature of the TLD and, where
appropriate, restricting who could register in that TLD. The proprietor could
do this without any involvement by ICANN. Would- be registrants in (say) .sports,
unhappy with the charter of that TLD, could choose to register in (say) .athletics
instead.
Charters are somewhat more problematic in a world in which the number
of TLDs is still tightly controlled. In a world in which there is a single
TLD directed at (say) banking, the chartering authority for .banc can exercise substantial
power in setting its rules for who may or may not register in that TLD. As
ICANN's designee to run the .banc TLD, it has quasi-official status. Because
of these factors, the chartering authority has the opportunity to engage in classic
restraint of trade. There are few sanctions available if the chartering authority
does not do a good job. ICANN may be called on to monitor its performance,
a job not especially well-suited to it..
It makes sense for ICANN to explore chartered
domains. But it should recognize the disadvantages of such domains if ICANN
empowers the chartering authority at the same time that the TLD is insulated from
meaningful inter-registry competition.
Q29: Are the first three principles outlined
in the second additional consensus point of WG-C's 17 April 2000 supplemental report
(quoted above) appropriate criteria for selecting TLDs to be introduced in the first
group?
The principles are appropriate, but they are widely misunderstood.
These criteria were carefully drafted. The first principle indicates that proposed
TLD strings should have some "significance," and contemplates that a TLD string will
typically have semantic meaning in some language. The second principle indicates
that TLDs may seek to enforce charters. It makes it equally clear, though,
that a TLD need not enforce a charter, and, indeed, need not have a charter.
The third principle states that because TLD strings "should not confuse Net users,"
TLDs should be "differentiated by the string" — that is, a TLD string should not
be confusingly similar to some other TLD string. It would be undesirable to
have both .com and .con in the name space, regardless of the purposes of the associated
TLDs.
These principles, correctly understood, are not a brief for chartered TLDs,
and they certainly do not preclude the creation of open TLDs. (Indeed, Working
Group C, which approved these principles, voted overwhelmingly to endorse the creation
of open TLDs.) The principles, indeed, are quite modest. They stand for
the propositions that TLD strings should generally have semantic meaning in some
language; that TLD strings should not be confusingly similar to each other; and that
some TLDs may have charters. That's it.
Q30: Do those principles preclude
the introduction of any new fully open TLDs?
See above.
* * *
Q32:
Should chartered TLDs be introduced according to a pre-defined system, or should
proposals be evaluated on an individualized basis?
The impulse to create
a "system" of TLDs is superficially attractive, but it should be resisted.
Such a top-down structure will inevitably be out of synch with the world it is meant
to mirror; even if it works at first, it will soon be obsolete. Serious study
of trademark categories makes it clear just how difficult it would be to devise a
useful, all-encompassing system of chartered TLDs. The name space will be better
served by a bottom-up system in which many TLDs are introduced in response to consumer
demand, and thrive or stagnate on the basis of consumer response, than by an official
category schema imposed from the top down.
* * *
Q34: Has the inventory of useful
and available domain names reached an unacceptably low level?
Yes.
While the number of random alphanumeric strings that can be used as an SLD is huge,
the number of strings with semantic meaning is much lower, and the number of attractive
and commercially useful domain names is much lower than that. Further, short
names are more commercially attractive than long ones. To appreciate this,
consider the secondary names market. Domain names fetch huge sums on the secondary
market. This reflects an overall scarcity of attractive and commercially useful names,
particularly in .com, which is seen as the premier Net address. The great virtue
of adding new open domains, assuming that the new domains' marketing and branding
is sufficient to establish them as meaningful competitors to .com, is that it would
sharply increase the number of short, meaningful, attractive domain names available
to individuals and firms on the Net. This would, in turn, drive down prices on the
secondary market.
Q35: Assuming it is important to increase the inventory of available
domain names, should that be done by adding TLDs that are not differentiated from
the present ones?
The best way to increase the inventory of available names is
to add open, unrestricted TLDs. These TLDs should be "differentiated" in the
sense that that term is used in Working Group C's principles: They should be identified
by TLD strings that are not confusingly similar to other TLD strings in use.
See my answer to Q29.
Q36: Should the formulation of policies for limited-purpose
TLDs be delegated to sponsoring organizations? In all cases or only in some?
See my answer to Q28.
* * *
Q41: Does the start up of a new TLD pose
additional risks to intellectual property rights that warrant additional protections?
No.
The start-up of a new TLD presents certain opportunities for cybersquatting and trademark
infringement; those risks are appropriately addressed via the UDRP. There is
no need for protections beyond the UDRP (and judicial proceedings under relevant
national law).
Q42: Should the protections afforded intellectual property
in the start-up phase of new TLDs differ depending on the type of TLD?
If the
UDRP were correctly applied, the likely result would be that no transfers would take
place in cases where the respondent registered a name in a TLD reserved for personal
or noncommercial domains. It might add useful clarity and certainty, thus,
if the ICANN provided that UDRP proceedings simply did not apply to names registered
in those domains.
Q43: Is the availability of the UDRP and court proceedings as
remedies for violations of enforceable legal rights an appropriate element of protection
of intellectual-property rights that should apply to all new TLDs? Are there any
other protections that should be made available in all new TLDs, regardless of their
type?
See above.
* * *
Q46: Is exclusion of names appearing on
a globally famous trademark list a workable method of protecting such marks from
infringement at the present time? Would an exclusion mechanism be appropriate in
the future?
No. As deliberations and discussion over the past year
have made clear, the creation of such a list would be monumentally difficult.
Nor does it make sense, in the absence of such a list, to create special "sunrise"
protections for *all* trademark holders. Under the latter approach, the exclusive
right to register "grass" during the sunrise period would go to trademark owners
including Grass Products, Inc. (who make aftershave), Grass Instrument Company
(who make scientific instruments) and Alfred Grass Ges. m.b.H. Metallwarenfabrik
(who make metal fittings for furniture). The exclusive right to register "computer"
during the sunrise period would go to trademark owners including Dunlop Olympic (Australia),
which uses "computer" as a trademark for socks. But a DNS in which a socks
manufacturer has a priority right to register the SLD "computers" across the range
of domains is a DNS governed by silly and arbitrary rules.
Q47: Should introduction
of new TLDs await completion of an evaluation of the operation of the UDRP and be
subject to a finding that the UDRP has been successful in meeting its objectives?
How long would such an evaluation likely take to complete?
No. Such
an evaluation would add unnecessary delay. There is little dispute that the
UDRP has provided trademark owners with a swift and inexpensive tool against cybersquatting.
*
* *
Q54: Should ICANN select the TLD labels, should they be proposed
by the applicants for new TLD registries, or should they be chosen by a consultative
process between the applicants and ICANN?
TLD labels should be proposed by the
applicants in the first instance.
Q55: Should there be minimum or maximum
length requirements for TLD codes? Are restrictions appropriate to avoid possible
future conflicts with ISO 3166-1 codes?
No length requirements are necessary;
registries can decide for themselves what TLD strings are most attractive.
Codes on the ISO 3166-1 reserved list should be excluded. (Alternatively, all
two-letter codes can be excluded.)
Q56: Should there be restrictions on the
types of TLD labels that are established (for example, a prohibition of country names)?
On
the one hand, it is undesirable for the name space to include TLD strings that are
inherently confusing (for example, .usgov for a site not controlled by the U.S. government).
This principle is treacherous, however, because reasonable people can differ widely
as to what categories of strings are confusing. An argument might be made that
TLD strings identical to country names would be confusing, because users would believe
that those TLDs were sponsored by national governments. I don't believe that
that would be an appreciable source of confusion.
* * *
Q58: How many
new TLDs of each type should be included in the initial introduction?
It would
be appropriate to introduce ten new TLDs in the initial deployment.
* * *
Q70:
How should ICANN evaluate the sufficiency of proposed intellectual property protections?
Registries
should not be proposing intellectual property protections; it is undesirable to put
registries in the position of policing intellectual property. Intellectual
property protections should be proposed and put in place by ICANN. As a general
matter, they should be limited to UDRP availability.