> So far as I know, consumers can edit the fields in their network control
> settings for most OSs to direct themselves to any DNS server willing to
> resolve names for them.
Actually it is a bit more complicated than that.
Relatively few user machines really point directly at any root server
system.
Rather, the settings on a host generally point to an intermediary server
that acts as a resolver. This is a good thing because the intermediary,
unless it is on one of those "operating systems" tends to run long enough
to build up a reasonable cache of resolved names (and hence helps the
system scale.)
These intermediary servers are the things that usually point to a root
sever system and these machines are generally run by an user's ISP or the
user's organization. Some of us (like me) run our own.
I'm now making a context switch and jumping into the notion of multiple
roots....
The key is to change our way of thinking about the DNS -- rather than
treating it as a core, fundamental service, it can be conceived of as a
service that can be offered by any number of providers and that users
chose among those offerings according to their needs.
We have an existance proof that this works -- there are a large number of
publishers of telephone directories. Some of these are published by the
phone companies, some are by folks who want to sell yellow pages(tm)
advertising, some are on CD-ROMs (hence the Supreme Court case on the
copyrightability of collections of facts), some are on web pages.
Yet, they all lead to usable answers. If any one of these didn't it would
lose market share and fade away.
The DNS can work the same way - we simply consider that there can be a
multiplicity of root systems, all operated independently, and all putting
into their "inventory" those TLDs that they think they can sell.
Because TLDs are now inventory, the root system operators will avoid those
TLDs that are troublesome, particularly those that are disputed.
That gets rid of the debate over how many and which TLDs to have -- the
competition of market economics decides.
Every root server system operator will try to trump its competitors by
having the most complete inventory of TLDs. This leads to an end
situation in which every root server system has all the viable TLDs, so
they are all essentially equivalent.
Then the distinction becomes that of value added services. And yes there
are value added services that a root server system can offer to its
customers. I've thought of a few that I mentioned - pointing to versions
of the TLDs that filter out porn sites, or offering free registration
services in your TLD to people who use your root system, etc etc.
As for fights over TLDs -- this approach says "let the contestants 'em
duke it out among themselves using standard economic and legal weapons."
We don't need any new bureaucracy to decide, we'll just let the law evolve
as it has over hundreds, if not thousands of years, through the resolution
of actual disputes between actual parties.
There are those who will wail and gnash their teeth and paint pictures of
doom, of internet instability, of internet non-connectivity.
To them I say: yea be of little faith in economics and competitive forces.
It is my assertion that multiple root systems, if recognized as legitimate
rather than routinely kicked and condemned, would quickly evolve into a
very stable system -- each root system would carry all the TLDs that
people want, otherwise the root system with the inadequate inventory would
failas customers migrate to places they like better.
And the TLDs will be the TLDs we expect. Any root server operator that
includes versions of TLD servers that don't return the answers that people
expect is a root server operator who is soon going to learn the phrase
"Would you like fries with that" at his/her new job.
In addition, the net would be in fact more stable with multiple root
systems -- we would have elminated a single point of failure. We have
seen what happens when a root zone is polluted, all kinds of bad things
can happen. With multiple roots, we have an easy and hot fallback for any
user that happens to encounter root pollution.
The Internet has run for some years now with some small-scale multiple
root systems operating in parallal to the one big "authoritative" root
system. I've used 'em. Nothing went wrong. I can attest from direct
experience that it it all works just fine.
One reason it that it hasn't taken off is that those who espouse this idea
have often been belittled and condemned and the idea is never allowed to
be discussed fully and without prejudice. I hope that that is no longer
the case.
Another reason is that there is a great deal of momentum behind the
current root system. That's why I've explored the thought that there can
actually be value-added differences that would create a reason for a user
(or, in most cases, the administration of whatever intermediary server is
being used) to switch.
Certainly, if ICANN is to truely try to say that it acts through voluntary
agreements and is not Internet Government, then there must be practical
alternatives. And for there to be alternatives in the DNS space, then it
must be clearly and unequivically recognized that other root systems then
the one under ICANN are legitimate, are to be permitted, and will be
freely allowed to flourish or wither without any oversight or control from
ICANN whatsoever.
--karl--
