Re: Shared domain correctness/quality considerations (was C

["Marty Modell" <ir001264@mindspring.com>]

Shared TLDs have in common the same problems that distributed 
replicated databases have in the commercial world.  Fine if they are 
read only, but a bear of a problem if multiple site updates are 
involved.

Most resolutions I know of rely on some central rbitartor site to 
carry the official update version which then replicates out to the 
field... barring that they rely on shadow updating.. the update is 
good until it is processed by some central site whcih resolves 
(somehow and how varies and is not relevent) duplications

Registries can perhaps work like state license plate agencies, I will 
sell you the special plate you want, subject to verification of 
availability) or you can have one from my assigned pool.

I get paid for the processing service, for whatever and I pay the 
central site for its services to me.

Does this work... yep... in about fifty states here in the US

> George Michaelson wrote:
> > 
> > What do you see as the minimum delay period between a name request and
> > lodgement that satisfies some kind of dither on fcfs to ensure legitemate
> > owner concerns can be met?
> 
> 	If we distinguish the Coordinating DataBase (CDB) 
> from DNS servers which are updated from it, then I'd suggest
> 	CDB: < 1 second
> 	DNS: ~ 1 day
> 
> 	I claim that anyone should be able to tell if there
> is a reasonable probability of being able to register a name (DNS),
> and that when actually registering it, the yes/no response
> should be immediate (CDB).
> 	To make this even stronger, I claim that the same
> immediate response should be possible to followup requests
> if the name is taken and the customer wants to try another.
> 
> 	I'll weaken it to admit that there is such a thing
> as the speed of light and the slow of database, and say that
> 3 seconds as a lower bound and 10 as an upper might be a good
> engineering target...
> 
> > Should there be a minimum hold time, and is
> > the size of a pool of iTLD admins a function of that time?
> 
> 	IFF you mean time that a name must remain invariant, I have no
> 	opinion.
> 
> 	IFF you mean time it will be held while filling in additional
> 	required information, I suspect an hour **might** be reasonable.
> 	This will vary with the customer data required (eg, 	
> 	authentication) and the minimum technical ability of the
> 	registry to provide/create/enable it.   We need a better
> 	spec before I'll go further.
> 	
> > 
> > What do you see as the maximum time between peer admins of a given iTLD
> > to re-sync data, such that the iTLD space is in some sense "dirty"? There
> > definately will be some period of dirtyness in an async world...
> 
> 	I see NS, RP and glue records being sent from the CDB
> 	to master dns servers within an hour of the registration
> 	completion, inter-server propagation on the order of a
> 	day and all other propagation as defined by TTLs.
> 	
> > 
> > What legal (or other) process enforces a party to honour  the requirement
> > that all iTLD are run as a global common? Given some peoples mail so far,
> > its reasonable to posit some set of bodies refusing to accept oversight
> > since they claim its illegitemate in the first place.
> 
> 	I do not wish to address this point here: I am
> 	of the opinion that it's a self-curing problem, but
> 	we should debate this in a different thread.
> > 
> > Who arbitrates over name removal and contention in any allocation where
> > one party would be inclined to permit it for a given iTLD and another iTLD
> > admin interprets policy as excluding the same name?
> 
> 	The customer (end usr, owner of the domain, etc) tells
> 	the parties what to do with its name.  The parties to
> 	any dispute are NOT registries. They are customers.
> 
> 	This does not consider the related question of whether
> 	to accept scatological terms. Someone needs to do a
> 	literature survey before rauising that question!
> > 
> > Apart from that, I think you've come closest to expressing goals here in
> > an appropriate way: Its not about making money. Its about global win.
> > 
> 
> 	Thank you for raising these questions!  In my copious spare
> time I'm collecting problems in hopes of overconstraining the
> specification problem (and yes, I did say **over** constraining)
> 
> --dave
> -- 
> David Collier-Brown,  | Always do right. This will gratify some people
> 185 Ellerslie Ave.,   | astonish the rest.        -- Mark Twain
> Willowdale, Ontario   | davecb@hobbes.ss.org, canada.sun.com
> N2M 1Y3. 416-223-8968 | http://java.science.yorku.ca/~davecb
> 
--  
Marty Modell                     e-mail: ir001264@mindspring.com 
      http://www.mindspring.com/~ir001264/Home.htm         
                  author of 
      A Professional's Guide to Systems Analysis
         Second Edition - McGraw Hill - 1996