> Would you suggest a "common root server" model (where, for example, all
> roots point to the A server as well as others) or a "mirrored root server"
> model (where, for example, A, B, and C root servers are identical and used
> for the common point) for the central control of the root?
The best way to answer you is suggest that you check out:
http://www.cavebear.com/cavebear/growl/
In the second second section of the current issue I describe what I'm
thinking of.
In essence it is completely distinct systems of roots, operated with no
imposed coordination except the enlightened self-interest of their
operators.
Those systems of root servers build their "inventory" of TLD pointers
according to what they think they can sell to their user/customer base.
Each root system operator selects which TLDs will be included and which
will not be included.
As I describe in the note, there are economic pressures which will drive
each root system to include all of the "viable" TLDs.
And there are value added services that can give a user reasons for
selecting one root server system over another. (Yes, there really are
value added services -- it was a big surprise to me, but, in fact, there
are several, although I only mention one in the note I mention above.)
As for the TLDs -- I'd leave it to the TLD operators to duke it out among
themselves using traditional legal and economic methods, and without any
top-down imposed regulatory structures. TLDs which are disputed are
likely to be considered "not viable" by the root operators.
The most interesting part is that there is not a thing that NTIA or ICANN
can do to stop the creation of multiple root systems. The technology is
in place and deployed. And there are already several existance proofs
that it works and that the net is not destabilized.
--karl--
(As an aside, there is utterly nothing special about an "A" server -- the
one used by the current roots is simply a legacy of operational
procedures, nothing more.)
