Routing Around A Single Point Of Failure
Date: Wednesday June 06 2001, @10:13AM
Topic: Alternate Roots

SimonHiggs writes "Here is how the Internet functions and why the Alternative Roots have sprung up around ICANN."

The Ideal World:

In the first illustration (figure 1), we start with a message from a user on one computer, and we want to send it to a recipient on a different computer.

Figure 1

As you can see, the message finds its way to the recipient computer, by the most direct route, which may often take it through intermediary computers along the way.

When things break down:

The next illustration (figure 2) shows the classic scenario of the Internet being subjected to irreparable damage to one of its nodes. The necessity still remains to get a message from one server to another. The system was designed to route packets via the most optimum available route.

Figure 2

Fortunately, the structure allows it to adapt quickly to damage, and we can route the message "the long way around" if necessary. As you can see, the message here gets through because there is no single point of failure.

ICANN and Alternative Roots:

Once again, here is an illustration (figure 3) with a single point of failure for the DNS root, surrounded by various Inclusive Roots. ICANN, now perceived by the Internet Community as failing to provide the necessary competition, has become a critical point of failure on the internet. Fortunately, as explained in the examples above, the Internet will route around a point of failure.

Figure 3

Vint Cerf, the new Chairman of ICANN, stated on CNN recently that no new TLDs have been introduced in the last 10 years, justifying ICANN is able to handle a very difficult and complex task. The statement wasn’t entirely true since there were 30 country-code TLDs added in 1995, 31 added in 1996, 47 added in 1997, 2 added in 1998, 1 in 1999. There have, in fact, been well over 100 ccTLDs added to the U.S. Government root in the last five years. There have, however, been no "generic" or gTLDs added in the last 10 years. The lack of free market access to the U.S. Government root, which is wholly managed by ICANN, means that the root system is not inclusive, but rather exclusionary, causing ICANN to be seen by the Internet as a point of failure.

In comparison, the ORSC Root has added all of the above ccTLDs, and over 100 additional gTLDs to what was the IANA root, making it 100% compatible with the current U.S. Government root, and yet serving an Inclusive Name Space at the same time.

This discussion has been archived. No new comments can be posted.
Routing Around A Single Point Of Failure | Log in/Create an Account | Top | 5 comments | Search Discussion
Click this button to post a comment to this story
The options below will change how the comments display
Check box to change your default comment view
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Re: Routing Around A Single Point Of Failure
by hta on Saturday June 09 2001, @11:11AM (#781)
User #2773 Info
Of course, Higgs is using a mesh structure as an arguments against a particular way of running a tree structure.
The UUCP addressing system (machine1!machine2!machine3!user) was mesh structured. It is also dead.
Rerouting (and routing, for that matter) requires a common way for nodes to agree upon identifiers for a destination.
And that is where Higgs' arguments break down.
[ Reply to This | Parent ]
Re: Routing Around A Single Point Of Failure
by markneill on Thursday June 07 2001, @05:02AM (#759)
User #2892 Info
They aren't intended to be generic codes, they are ccTLD's. The countries that use them have marketed them as gTLD's.

Vint is wrong, but not for the reason you post. He's wrong because there have been TLD's added in the last 5 years, in bursts, with no illl effects.

However, this article is also wrong. The destruction of the root servers will not be "routed around" as the author suggests. In routing, everyone knows aboute everyone else, in a roundabout way, or at least knows about all of their accessible neighbors.

DNS is not normally configured this way. A DNS server has a root stub, and pointers to a number of root servers. These root servers are, for the most part, the ROOT-SERVERS.NET servers, which are controlled by ICANN now.

If those servers go away, there is noting else to route them too, without manually reconfiguring thousands of DNS servers.

It can be corrected, but it is not self-healing.
[ Reply to This | Parent ]
  • 2 replies beneath your current threshold.

  • This article comes from ICANNWatch

    The URL for this story is: