May 30, 2005

An Infrastructure TLD - Avoiding the Side Effects of Today's .net

I've mentioned before that there is something special about the .net top level domain - in particular .net is the place where the legacy root DNS servers and most of the TLD servers are to be found.  Thus, if .net were to wobble there is more than a strong chance that the DNS root and other TLDs would also begin to wobble.  This kind of cross-dependency is something that A) is a risk to overall internet stability and B) is something that ICANN seems utterly unable to perceive.

So I ask this simple question: Why can't the domain names of the legacy root servers and TLD servers be moved to a new global infrastructure top level domain?  Such a new TLD would be intrinsically much more stable than .net.  In fact because the size would be small, a new infrastructure-only TLD could be readily cached and replicated, thus providing much more resiliency against attack and could be recovered much more quickly than .net should an attack be successful.

This new TLD should be used only for machines that provide services in support of DNS on a global basis (with the proviso that any server that delivers a TLD zone for any TLD, whether that TLD is ICANN approved or not, should be considered "global infrastructure".)

For the moment let's call this new TLD "q8m", which is a short phrase without any annoying semantics (I hope).

Thus this infrastructure TLD would contain delegations for things like "root-servers.q8m" and "tld-servers.q8m" to replace the existing "root-servers.net" and "tld-servers.net".

Anyone who wants to establish a group of infrastructure servers would register for a delegation in this this infrastructure TLD.  The registration agreement would require that the registrant police the use of the delegation so that the resource records found via that delegation are all present for the exclusive purpose of providing infrastructure services.

In order to discourage spurious thrashing of the contents of this infrastructure TLD there should be a steeply ramped fee schedule for updates.  The first 4 per year should be inexpensive ($25) but after that the fee should quickly ramp up to at least $100 per update.

Were this kind of infrastructure TLD to be established, much of the special nature of .net would be eliminated; a failure of .net would not then have the kind of destructive repercussions onto other parts of the internet that is now the case.

Posted by karl at May 30, 2005 11:54 AM