Karl,
Very good document, indeed.
Let me state my POV on subject #2 (I have objections also on #1, but as this
has been discussed several times, all readers know where our positions
differ, and there's no need to bore them with another thread).
I have nothing in principle against multiple roots, I just fail to
understand how this could be a better system than a common root under public
trust.
I find the analogy with the phone system (as you present it) not fully
applicable, as the phone number is a "key" in the system, and therefore
unique due to the way that the system is built, while the domain name is an
"attribute" of the unique key (the IP address), and therefore could be
duplicated.
The problem comes from the fact that while you use the "key" in the former
system (you call the phone number) you currently use the "attribute" in the
latter (you type in the domain name, not the address).
The way I see it, either you enforce uniqueness of the TLD across the roots
(and therefore you have a logically unique root which is the logical union
of the roots), or you may well have the TLD resolving differently in two
different roots. In other words, if you have two systems serving .web (IOD
and CORE ?), you may have two different instances of
"i_am_the_best_on_the.web", if one belongs to a registrant with IOD, the
other with CORE.
If I understand your position, you claim that the market will sort out who
will survive, in the sense that service providers will point to the one that
will offer more added value (BTW: to them, not necessarily to the end user).
This means that you will find the one or the other according to how your
system is set.
It is like saying (in the telephone system analogy) that by calling +1 831
423 8585 you may reach "cavebear" or a subscriber to a different provider,
according to how your telephone is set. This is far worse that being listed
in one or the other directory. For instance, end users could get different
two directories, and call numbers referenced in both without any change in
the setting of the telephone. This is not true in the ISP-chosen root system
(unless you don't manipulate your own computer, which very few people will
do).
Of course, as E-Commerce grows, companies are investing more and more in
this new tool, and the last thing that they want is to have their efforts
jeopardized by the dependency from a setup. And this will be independent
from their performance, but the risk factor is related to the performance
(as perceived by the service providers) of the Registry they belong to. Up
to the point that they may need to change Registry, and therefore often even
domain name, if their first choice is already taken in the other Registry.
I believe that commercial organizations see this as a nightmare.
BTW, this is the reason why commercial interests are supporting ICANN: they
have nothing to do with CORE, to whom they are completely incorrelated, but
just feel that between the ethics of holding closed meetings and the
financial risk involved with limited visibility on the Net they consistenly
choose the lesser evil.
All what we need, as a first step, is more competition (i.e. new TLDs,
allocated to different players), not to rethink the whole thing over again,
or go into unexplored territory (where what is unexplored is not the
technical feasibility of the multiple roots, which is proven, but its
effectiveness on large scale numbers).
I will be glad to stand corrected if I missed something of the point you
make and/or if you find a hole in my argumentation, as not only I am not a
lawyer, but I'm not a guru of the IP addressing either ;>).
Best regards
Roberto
P.S.: I agree with you on point #3
