Re: [wg-c] Re: IP/TM Concerns & New GTLDs

[Kent Crispin <kent@songbird.com>]

On Mon, Aug 02, 1999 at 02:10:44AM -0700, Roeland M.J. Meyer wrote:
[...]

> As far as speed of deployment, where will the CASH come from, the ICANN?
> Even at five registries, we are talking about over $20MUS in costs. If
> they have to be non-profit, with a dictated business model, they will
> NEVER come into being.
>
> In a prior message, I discussed costs of registry operations.

Your figures have a number of flaws.  They assume a from-scratch
development of a stand-alone proprietary registry.  In fact, of
course, the most likely case is that the registry would be an
incremental addition to an already existing facility with 24x7 NOC
etc.

Furthermore, even given the pure stand-alone development, your
figures are highly inflated.  I agree that in general people costs
are frequently underestimated, but you go very, very far in the other
direction...

[...]

> > > And it's no good to say that you're only going to give it
> > that monopoly
> > > for a year or two. The advantage such a registry would have over its
> > > rivals, given the pent up demand, would be significant.
> >
> > Only if you assume that the registry operator "owns" the TLD.  If it
> > doesn't, and it is run as a non-profit, then TLDs can be re-assigned
> > to new registries any time.  Recall that requiring the registry to be
> > non-profit does not require the registry operator to be non-profit
> > -- CORE is a non-profit, Emergent is a for profit.
> 
> and the ONLY successful registry, NSI, is a for-profit.

That is nonsense, of course.  Nominet and the German registries are
the largest European registries.  Unfortunately, I don't read German,
but Nominet published their annual report on their web site -- as a
non-profit they have the embarassing problem of swimming in cash. 

> > It's important to realize that there are essentially no economies of
> > scale to worry about.  I could run a significant registry from my pc
> > at home.  This is technical fact.  It isn't widely known, but one of
> > the .com/.net/.org TLD servers actually is just a PC -- true, it's a
> > multiheaded Pentium server with a Gig of memory and lots of fast disk
> > and really good network connectivity.  But it's a PC.  And that's the
> > DNS server, which gets ~2000 hits/sec.  The registry database, as has
> > been mentioned before, gets orders of magnitude less traffic than the
> > dns server.
> 
> Yes, and who is running it and what is the cost of that data center
> context?

The server is maintained in an educational institution, I believe; 
most of the cost of the data center operations and infrastructure 
are already covered.

> What is the fully-burdened cost of that resource? You are
> vastly understating that server.

Couple of FTEs.  Most of the work is done by one person.

> Yes, the whole thing could be run on
> two or three racks. But the amount of infrastructure work is awesome,
> especially for a proper service offering. The last estimate I put
> together priced it out at $2.9M just for the development of the registry
> software, not to mention on-going license fees and maintenance.

Willie Black developed the registry software for Nominet in his spare
time; the original contract for the CORE SRS was $700,000, but the
CORE SRS was designed to handle the .com load, and that included all
development costs, not just software.

[...]


> A sanitized version of that offering is available at
> <http://download.dnso.net/dox/politicks/IDNRS.pdf>

In general, I agree that people frequently underestimate labor costs,
but you would do very well indeed, if you could sucker someone into
that.  Eg: 2856 hours for a data modeler -- that's 1.5 years,
full-time.  This is *not* a complicated database... 

> > The registry part of the registration business (as opposed to the
> > registrar part) is *insignificant*, relative to the registrar
> > component.  There are over a thousand Nominet registrars feeding a
> > single registry.
> 
> What sort of commission are they paying?

Starting September they will pay ~ $4.50/year to register a name.  
It's not a commission; it's more a wholesale/retail arrangement.

[...]

> > You are vastly overestimating the value of competition at the
> > registry level.
> 
> As I think that you are totaly mis-understanding how channel marketing
> works. Channel marketing is not competition. Only registries are true
> competition. Channel markets are only a means of leveraging the work of
> others, for the benefit of the channel operator.

Proof by buzzwords?

We have concrete examples of WORKING shared registry systems.  They
are doing quite well; the registrars in fact compete.  These are
facts, not buzzword-laden theories. 

[...]

-- 
Kent Crispin                               "Do good, and you'll be
kent@songbird.com                           lonesome." -- Mark Twain