IP: More Re: Basket and eggs, was Re: ICANN and IBM

[David Farber <farber () cis upenn edu>]

> Date: Sat, 25 Sep 1999 23:52:37 -0400
> To: Ed Gerck <egerck () NMA COM>
> From: Dave Crocker <dcrocker () brandenburg com>
>
>
> At 02:24 PM 9/25/99 , Ed Gerck wrote:
> > The above is false in totum because its central argument, that "a 
> > central third
> > party was needed to manage the assignment of domain names", is false.  The
> > DNS was created exactly to avoid that, which was the case of the 
> > old HOSTS.TXT
>
> Right and wrong.  Right that subordinate portions of the tree are 
> delegated.  Wrong in that the scheme is inherently "central" for 
> each node in the name hierarchy, including the root and that those 
> delegations are under the authority of the next higher node, all the 
> way up to the root.
>
> > file. The mechanism of zone delegation in DNS allows the assignment of domain
> > names to be fully decentralized, and yet globally unique and meaningful.
>
> It allows a given node to be DELEGATED by the next higher node's 
> authority.  The delegation is to a SINGLE authority for this 
> next-lower node.  These are constructs of centralized authority, 
> albeit with controlled delegation to improve administrative and 
> operational scaling.
>
> > The only object  that needs to be centralized in the DNS is a 
> > common reference for
> > *name resolution* -- which is the root name-server.
>
> Each node in the hierarchy has the same need for centralized 
> control, not just the root.
>
> > To take over *name assignment*
> > by confusing it with name resolution is an over-reaching action, 
> > unjustified by
>
> The current scheme for centralized administrative control -- albeit 
> with delegation UNDER that control -- of both assignment 
> (registration of names) and operation (DNS primaries for each level 
> in the hierarchy) has been present ever since the start of the DNS.
>
> The confusion is in misunderstanding why each was designed to their 
> current form and in believing that they can or must be changed, 
> absent detailed, reviewed and approved specification of the changes.
>
> > any technical argument even in the current DNS (i.e., even 
> > outcounting eventual but
> > certain technological improvements and paradigm changes in the DNS).  Which
>
> This sort of language appears to demand policy changes in advance of 
> the unnamed technical "advances".  That's more that a bit risky for 
> a critical piece of operational infrastructure.
>
> > Further, by placing the decisions of network address assignment (IP 
> > numbers) together
> > with DNS matters in one basket, you are uniting what is, by design, 
> > separate. The
>
> This nicely ignores that the functions have been co-located since 
> their inception.  While it is fine to explore the possible benefits 
> of decoupling them, it is irresponsible to demand changes to an 
> existing and highly effective system, absent that debate and absent 
> a resolution to that debate, making compulsory the split.  (Minor 
> items like the redundant costs and administrative overhead that 
> would be incurred are examples of counter-points to the unlisted 
> points in favor of a split.)
>
> > So, IBM and you are basing your course of action on reasons which 
> > are unreasonable.
>
> And you, NMA, are basing yours on premature, poorly understood 
> demands which are more likely to break the net than to create any 
> benefit.
>
> > 1. not market-accountable (a non-profit with no measurable market 
> > value in stock
> > 2. not community-accountable (no elections),
> > 3. not anti-trust accountable (it is a government appointed company, and the
> > 4. not legally-accountable (has no assets to be put at stake; has no owner),
> > 5. allows registrars and registries to be also non-accountable (a 
> > TLD registry can
> > 5. makes end-users the only entity accountable in the entire 
> > system, the only ones
>
> This litany is sorry not just for its being repeated constantly, at 
> the expense of the facts, but for the astonishing aspect that it is 
> taken seriously.
>
> > to be necessary for a central control system -- the one basket. 
> > The fact, however,
> > that no one needs to control name assignment is however a reason for the eggs
> > not to be put into what has become a trap.  And, the more solid and 
> > unified the
> > trap is, the less reasons for any egg to be put there. Presumably, 
> > not even IBM.
>
> Since there need not be central control, you are invited to create 
> your own system and leave the current one to evolve rather than be 
> distorted to your model.  The Internet is mostly tolerant of 
> parallel efforts.
>
> In fact there have already been a number of efforts to replace the 
> current DNS system, though none has gained popularity with more than 
> 1/2 of one percent of the Internet's users.  Further, no alternative 
> administrative or operational scheme has been put forward to 
> Internet technical or operational standards bodies, never mind 
> approved by them.
>
> > And, let's be candid. The issue is Internet and information control.  Yes,
>
> Well that is certainly what a few people have turned it into, rather 
> than permitting a natural evolution of a system that functioned 
> quite well since its inception, going through roughly 8 orders of 
> magnitude of growth.  Those who turned this into a question of 
> control have nicely stifled further growth.
>
> d/
>
> =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> Dave Crocker                                         Tel: +1 408 246 8253
> Brandenburg Consulting                               Fax: +1 408 273 6464
> 675 Spruce Drive                             <http://www.brandenburg.com>
> Sunnyvale, CA 94086 USA                 <mailto:dcrocker () brandenburg com>