Comments of Dr. Milton Mueller, Associate Professor
Syracuse
University School of Information Studies
Prefatory comments:ICANN
is to be commended for preparing a thorough set of questions regarding policy for
new TLDs. Certain aspects of the questions and the accompanying statement are troubling,
however.
First, ICANN needs to define more precisely what it means by "stability."
As best as I can discern, the word is used to denote "any problem that might occur
anywhere in the Internet." This overly broad definition has the ill effect of encouraging
ICANN to insert itself into too many areas where it does not belong. "Stability"
is normally used to denote technical and operational stability of the DNS. If ICANN
is using the term to mean something else, it should offer a precise definition and
ground it in the White Paper.
Second, the ICANN call (section A) states, "Introducing
new TLDs implies a change in the overall structure of the DNS." That comment reveals
a surprising lack of knowledge of the DNS protocol and its structure. The DNS is
a hierarchical name space. While the business implications of adding TLDs are (in
some cases) significant, the technical implications are not much different than registering
a second-level domain name; indeed, there are many second-level domains that generate
more traffic than top-level domains. Adding new top-level domains consists of adding
a few lines of text into the root zone file that define the new TLD string and provide
addresses for at least two name servers capable of resolving names under that TLD.
Adding TLDs is not a change in the overall structure of DNS. It is an ordinary implementation
of DNS.
Third, the report also states that we have "no experience" with the addition
of new TLDs to the root. This is plainly false. Over 30 new TLDs – country codes
-- were added each year to the root during the mid-1990s. Technically and operationally,
there is no difference between ccTLDs and gTLDs. In the root zone file, all TLDs
are simply strings of text associated with the IP addresses of name servers. The
computers responding to queries of the root zone are completely unaware of whether
the TLD queried stands for a country or something else. It just looks up the name
and returns the record. There are no new operational or technical issues posed by
adding TLDs to the root. There are, of course, important resource allocation decisions
to be made regarding who gets them. One does not clarify this issue by falsely characterizing
it as a step into uncharted technical territory or by raising unwarranted fears about
the stability of the Internet.
On to the specfic questions:
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?
There are no protocol or standardization issues
raised by the addition of new top level domains per se. As noted above, it is a simple
implementation of DNS.
Q2: What stability concerns are associated with the initial
phases of registration within the TLD?
None. To put this in perspective, the com
zone is currently increasing by over 35,000 entries every day of every week. The
idea that any one of the ten new TLD registries would instantly attract that large
a stream of new registrations is not very likely. It is far more likely that new
registries will have to work very hard for their business. But if it did happen,
a registry with the proper facilities would be able to handle it. Indeed, it seems
more stable to distribute the growing load of DNS entries over multiple new registries
than to concentrate it on a single point of failure.
At any rate, a particular
registry’s quality of service is not an issue for the root manager. The stability
of the Internet itself is not threatened by a local failure, anymore than Internet
stability is affected when a local ISP’s dial up lines are all busy or a web site
receives too many hits to handle. Registries that are unable to handle the initial
phases of registration will lose business. If ICANN is wise and introduces enough
new registries to ensure competition, consumers will have plenty of other choices
and the problem will be confined to the registry.
Q4: Would these stability concerns
be magnified by introducing a large number of TLDs at once?
No. The DNS is a robust
protocol that has proven its scalability over the long term. The consensus point
reached by Working Group C was that a maximum of 10 new TLDs should be added. Unless
the number exceeds what is as yet an unknown threshold, but is certainly well over
a thousand entries in the top-level zone, there are no stability issues whatsoever
raised by adding new TLDs.
We know that NSI put the root zone on the same server
as the zone files for dot com for many years. During that period, the com zone file
was quite large, containing over a million entries. The developers of DNS such as
Paul Mockapetris and other technical experts such as Paul Vixie have made it clear
publicly that there are no stability issues raised by adding 10 new TLDs to the root.
Q5:
Are there any practical means of reversing the introduction of a significant new
TLD once it goes into operation?
Of course. You can take it out of the root.
If by "practical" you mean "costless and painless," then the answer is probably
no. The real issue is who bears the costs of the "reversal" and how the liability
gets worked out.
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?
Given
the switching costs associated with the use of domain names, this is not a commercially
viable way to introduce a TLD, so any activity within that domain would not be a
"trial" of anything except some people’s willingness to gamble.