[governance] How many TLDs?

Thomas Narten narten at us.ibm.com
Tue Mar 4 09:26:33 EST 2008


Karl Auerbach <karl at cavebear.com> writes:

> Sylvia Caras wrote:

> > For me, a central part of the top level domain conversation is how
> > many might there be.

IMO, this question misses the point. Clearly, if we could have an
infinite number of TLDs, there would be an almost infinite demand for
them. One need only look at the number of registrations in .com. Or to
think about how many vanity TLDs people might want to have.

It all comes to down to cost. If the cost to own/run a TLD is low
enough (say under $50/year), the demand could easily be in the tens of
millions.

But just because we _could_ support millions of TLDs and because there
might demand for millions of TLDs, doesn't mean we should or that we
need to.

>  From a technical perspective the technical burden of serving a root 
> zone to the net is equivalent serving a TLD zone.

This is true, in a narrow, "technical" sense, but it is misleading, as
it suggests there is no reason not to do so.

> Today the .com zone runs with roughly 70,000,000 names.  .com runs
> reliably and with an acceptable administrative error rate.  Verisign
> has done a good technical job.

And, if one were to look at the infrastructure needed to support the
.com domain (in terms of bandwidth to the servers to support typical
query rates, redundancy, hardware, ability to do frequent updates to
the zone, 7x24 operational support, over-provisioning for attack
mitigation, etc., etc.) there is a very significant cost associated
with running such a service.

Then look at the root zone. One of the nice things about the current
root zone is that it is (relatively) small, and the zone does not
change frequently (currently twice a day I believe). That means that
the capital and operational costs of operating a root server are
relatively small (compared with .com).

This is a good thing, since the root zone is not operated by single
monolitic organization. Indeed, the funding model for running the root
zone is rather unique. No one pays the root operators to provide a
service. Each operator pays for it through their own means. It would
be an interesting exercise to think about the implications of
(effectively) forcing them to upgrade their support capacity from its
current state to something on the scale of supporting .com.

In addition, since the root is in some sense a single point of
failure, replicating root servers is a very good thing to do in
general. With the use of anycast technology, we now have over 100 root
servers. That number should and will go up.

Indeed, so long as the root zone stays small, just about anyone could
run their own root server if they wanted to.

But, if each of those replicated root servers had to have an
infrastructure behind it equivalent to that needed to support .com,
that doesn't come for free. Someone has to pay for it, and it will
likely mean that some sites that would like to run their own copy of
the root could not do so. Think of less developed regions of the
world. Would this be a Good Thing? I think not.

All that said, I am NOT arguing that the root needs to be kept at its
current size. It could easily be expanded. Rough rules of thumb that I
have heard people suggest indicate that increasing the current size of
the root zone by two orders of magnitude could be done without any
serious issue/change to the current operations of the root zone. So, I
don't have any (technical) worries about adding tens or hundreds of
new TLDs per year, maybe more as we gain more experience with the
process (and with such difficult topics as how to deal with TLD
business failures).

But I am opposed to gearing towards millions of TLDs. That is the
wrong target to aim for.  The DNS works precisely because it is
structured as a tree, so that load is (by design) distributed widely
and quickly away from a central "root". Adding millions of TLDs to the
root zone undermines that fundamental design property, and I see
mostly downsides with little compelling benefit in doing so.

Just because we could in theory support millions of TLDs doesn't mean
we need to or even should.

Thomas
____________________________________________________________
You received this message as a subscriber on the list:
     governance at lists.cpsr.org
To be removed from the list, send any message to:
     governance-unsubscribe at lists.cpsr.org

For all list information and functions, see:
     http://lists.cpsr.org/lists/info/governance



More information about the Governance mailing list