Re: [ga] More on Sitefinder suspension

[Karl Auerbach <karl@xxxxxxxxxxxx>]

On Mon, 29 Sep 2003, Stephane Bortzmeyer wrote:

> > For example, consider the IAB/ICANN statement about competing roots
> > - those reports claimed that the sky would fall, the seas would
> > boil, and hell would freeze if competing roots were allowed.
> 
> This is not true. Anyone can check from:
> 
> http://www.icann.org/icp/icp-3.htm

I did just recheck, and by doing so I reaffirmed my opinion that ICANN's
ICP-3 is a work that is essentially non-technical. It asserts, like an
11th commandment, that there shall be one and only one catholic root but
does not really dig into the root (pun intended) causes.

I found the assertions regarding a global and uniform namespace to be
simplistic.

My own analysis of global/uniform name spaces may be found at
http://www.cavebear.com/rw/nrc_presentation_july_11_2001.ppt (The part
about global and uniform name spaces starts on slide #5)

> that ICANN or IAB never said so.

I was sitting right next to Stuart Lynn at a public ICANN board meeting
when he enunciated it as official ICANN policy.  So I can say from my own
personal knowledge that ICANN did in fact say "so". ;-)  (ICANN's ICP-3
was never adopted by the board of directors, so there is some ground to
argue that it was, and remains, merely Stuart Lynn's personal opinion.)

> > Yet competing roots have been here and working for years and years -
> > and the net still works.
> 
> Of course, "competing roots" are just a small set of local experiments
> used by the maintainer and a few pals. No wonder it works: almost
> nobody uses them.

I've been using 'em for my companies and my personal use for more than the
lifetime of ICANN.  I've had a questioning eye, I've been looking for
problems, and I've moved a non-trivial amount of traffic and cases through
the system over that time.  And during that time I have had exactly one
problem - the fact that ICANN introduced an .biz that was inconsistent
with the version that had been established and made operational well
before the ICANN adopted one was even suggested.

What is being missed is that the risk is not that of competing roots, but
rather, inconsistency.

Several years ago I wrote the following points.  These are intended to
focus the dissusion onto what I believe is the real issue - consistency - 
rather than the artificial issue, competing roots.

  There appear to be three distinct cases to describe the events that
  transpire when users of the Internet use different DNS roots.

  In order to simplify things let me adopt some simple terminology:  "Root-D"
  stands for the dominant (NTIA controlled) DNS root - this is the one that
  serves the vast majority of Internet users.  "Root-X" stands for any of the
  other root systems.

  The three cases seem to be:

  A) Root-D and Root-X have identical contents.

  Most people consider Case-A to be essentially a mirroring situation and
  reasonably benign.

  B) Root-X has more top-level domains than does Root-D but for those TLDs in
  common, the contents are identical.

  Case B represents the situation that obtains today between the NTIA
  controlled root and the other root systems.

  C) Root-X and Root-D contain at least one top-level domain with the same name
  but with different contents.

  Case C represents a situation that may readily occur and that most people
  consider pathological.  The real issue here is not a technical one.  Rather
  the question is what are the most effective means to either prevent the
  situation from arising at all or to create pressures that work so that
  these situations tend to occur only in somewhat private backwater areas of
  the net.

 
> I use Open-RSC at home because I love to play with new toys. Sure, it
> works. It directs me to the right name server, except for a few ccTLD
> that are not very important to me. Exactly like ICANN does. Read
> again: exactly like ICANN does.

Yes, which is exactly what a competing root is supposed to do - to direct 
one to the exact same TLD servers that one would get to via that catholic 
roots.  And that raises the question - since the information that comes 
back is the same, what's the fuss about?

> There is no reason to use a competing root, they provide exactly the
> same service.

There are, in fact, reasons to have additional roots - not the least is
one situation that I've had to endure several times, the loss of all
connectivity to "the outside" due to natural disaster.  The ability to
establish a local root is a very important tool for getting communications
up and running pending reconnect to the larger net.

Another reason, one that is diminishing due to the use of anycast for the 
catholic roots, is to add local root service capacity.

There is also the issue that is raised by the Verisign wildcard situation
- what prohibitions should exist on private acts on the net?

> All the other dummy TLDs that they are supposed to allow
> either does not work at all (most TLD in Open-RSC have not *one* name
> server running) or are useless since they never appear in email
> addresses, in HTML links, etc.

There is no doubt that there is a critical mass issue.  But that's a 
marketing/sales issue, not a technical one.

The ARPAnet 1974 was "useless" as viewed by most of the folks who were
around at that time.

In 1995 few people had ever seen URL's, but that didn't mean that they
were useless.

> Saying that "competing roots" work is exactly like saying that true
> direct democracy works because you run your commune that way: it does
> not mean it can work for the whole Internet.

Scaling is an issue - but there are two points to consider:

1. There is nothing intrinsic about the catholic root that makes it scale
better than competing roots.  Should Microsoft or AOL, or anybody with
some money to spend, wish to deploy their own roots, they could easily
deploy a set of servers with capacity that would be amazing.

2. I see few republics condemn communities who do have managed to 
establish direct democratic systems.  Yet the catholic root community 
seems to take every chance to deamonize the even the concept of competing 
roots.

		--karl--