ietf
[Top] [All Lists]

Re: Split the IANA functions?

2014-01-07 10:29:26
On Tue, Jan 7, 2014 at 9:44 AM, Andrew Sullivan 
<ajs(_at_)anvilwalrusden(_dot_)com>wrote:

On Tue, Jan 07, 2014 at 07:02:34AM +0100, Eliot Lear wrote:
That there is a single root is both a technical AND a
political decision.

I have to disagree with this.  The uniqueness of the root is not a
political decision.  It's a fact of mathematics.  DNS is a tree.
There's one root.


It didn't have to be a tree, it could have been something else, we could
still change it.

You might think that there is no other possible technical solution but that
is a failure of imagination rather than a fact. It might even be the best
technical solution but the definition of 'best' is with respect to
requirements.


A lot of the 'technical' arguments that are made against demands made by
Russia, China, etc. are actually technical choices and it undermines our
position when people falsely claim that there is no alternative.

I am currently at a conference where Fadi Chehadé just talked about Russia
asking to host a root server and he led the audience to believe that there
could only be 13 servers. So I had to correct him and point out that Russia
already has a DNS server due to anycast and what we are actually arguing
about is a root service, an abstract construct.

There is of course no technical reason that the number of services is
limited to 13. We can have as many as there are institutions willing to run
them at an acceptable level of reliability.

Putin's technical advisor is probably telling Putin that the westerner's
claims of technical limitations are bogus and he would be right.


The argument against more root servers is political, not technical and the
politics are rather more complex than just 'should Russia have one'.

At the moment the root is small and the load relatively light. In fact the
legitimate load is essentially zero. The root zone could probably be
distributed as a flat file, it isn't large, it doesn't change very much.
Certainly it is nothing like .com. The only real issue in root server
management is dealing with the load from the DoS attacks which are awesome
in scale.

The main consequence of expanding the root zone is that the root zone
operators would constrain future development of the DNS. The root zone is
going to grow with the new TLDs. When the number of TLDs reaches a certain
point there will be a tipping point and there will be pressure to open up.

Having a hundred root operators would greatly constrain those changes, But
that might in part be why Russia and China want a seat at that table

This is not to say that the single root doesn't have additional
political implications.  But I will not concede that this is some
contingent fact of the DNS that could be otherwise.  The particular
root we have could be different.  That there is a single root could
not.


Uniformity of the name space is the essential criteria. The Web can't work
on UUCP bang path addressing.

Describing this as having a single root conflates a large number of issues
and essentially commits to a particular conclusion. A uniform namespace is
a requirement, a single 'root' is not.



-- 
Website: http://hallambaker.com/