To:
"'Stephane Bortzmeyer'" <bortzmeyer@nic.fr>
Cc:
ietf-provreg@cafax.se
From:
"Hollenbeck, Scott" <shollenbeck@verisign.com>
Date:
Tue, 12 Nov 2002 07:27:53 -0500
Sender:
owner-ietf-provreg@cafax.se
Subject:
RE: Is is mandatory for an object to belong to a registrar?
> -----Original Message----- > From: 'Stephane Bortzmeyer' [mailto:bortzmeyer@nic.fr] > Sent: Tuesday, November 12, 2002 5:01 AM > To: Hollenbeck, Scott > Cc: 'Stephane Bortzmeyer'; ietf-provreg@cafax.se > Subject: Re: Is is mandatory for an object to belong to a registrar? > > > On Fri, Nov 08, 2002 at 08:44:06AM -0500, > Hollenbeck, Scott <shollenbeck@verisign.com> wrote > a message of 29 lines which said: > > > I deliberately tried to stay away from the concept of ownership > > OK, but replace "owner" and "owns" in my message by "sponsor" and > "sponsors" and I have the same concern. If I want to implement a > registry where contacts are not owned/sponsored/managed by registrars, > is it possible with EPP? I also don't use the term "registrar" anywhere in the documents. "Client" is used instead, and a client can be anything that talks to the server. > > "manage", etc. seemed like more neutral terms. I don't > think you'll find a > > single use of the word "own" to describe the client-object > relationship in > > the specs > > Right, but not the point. Sorry, but I think it's very relevent. You suggested that registrars "own" objects, and I'm saying that they don't. > > The answer to your question is "yes". The current protocol > specifications > > allow implementation of a model where contacts can not be > transferred. > > I was more specific: a model where contacts are not tied to a > registrar. For instance, <info> should not mandate: > > - A <contact:clID> element that contains the identifier of the > sponsoring client. > > (There is a <contact:crID> element if you want historical information > about who created a contact.) The model is one where objects are managed by some client. If you're asking if the <contact:clID> is optional in an <info> response, it's not. If you're suggesting that it should be, I would disagree. Someone or something has to be the authoritative management agent for an object. Now, if you're asking if that agent has to be a registrar the answer is "no". One might develop a model where the agent is the server (or a registry operator to use the registry-registrar terms you've used) itself. -Scott-