To:
Patrik F$BgM(B
tstr$B‹N(B
<paf@cisco.com>
Cc:
ietf-provreg@cafax.se
From:
Takeshi Saigoh <Ken@REACTO.com>
Date:
Thu, 08 Mar 2001 15:35:16 +0900
In-Reply-To:
<p0510015bb6cc52dce757@[10.49.160.101]>
Sender:
owner-ietf-provreg@cafax.se
Subject:
RE: Unique handle generation
Patrik, Our IOJ system assigns a contact handle by a similar way. ie: "NIC-<REGISTRAR ID>-<LOCAL USER ID>" This way functions well in a usual case. However, when a registrar stops service, this way brings a user confusion. If it adopts this way, the database which manages all handles is required (UWhois??). Of course, it needs to follow correspondence of an old handle and a new handle... Ken -----Original Message----- To: "Hollenbeck, Scott" <shollenbeck@verisign.com>,"'George Belotsky'" <george@register.com> From: Patrik F$BgM(Btstr$B‹N(BSubject: RE: Unique handle generation Cc: ietf-provreg@cafax.se Content-Type: text/plain; charset="us-ascii" ; format="flowed" Sender: owner-ietf-provreg@cafax.se Precedence: bulk Status: At 15.38 -0500 01-03-07, Hollenbeck, Scott wrote: >George, > >Maybe I'm missing something, but what's not unique about an e-mail address? A handle is a globally unique identifier of a record in a database. The handle can normally be used for two things: - Uniqueness - Locality In some cases when one design a handle, one have to choose one of these functions before the other, i.e. prioritize. Each record need a handle. A person change email address, but the handle of the record which contain the email address should stay the same -- and have as long life as possible. The routing registries today use handles which are of the form <LOCAL HANDLE>-<SERVER HANDLE> Example: PAFA1-RIPE The local handle have to be unique within the server, and the server handles unique by itself. The combination of local handle and server handle will because of this be globally unique. A record can if these kind of handles are used refer to objects in a different registry -- which I claim is a good thing. If we have a registry of all server handles, we can even locate the record given the server handle -- which I claim is a good thing. A record can though NOT with this design move from one registry to another, and I claim that is not needed. With registry I mean for example the three RIR we have today, or one TLD registry (including registrars). This means that all registrars for the same TLD have to use the same server handle, and unique local handles between themselves for this scheme to work. If not, if one transfer a domain (for example) from one registrar to another, the handle will change. This is a bad thing. So, my proposal is simply that the local handle is allocated / generated by the registry for each object, the server handle is registered somewhere (I have a draft which is on it's way out...but I missed the deadline) and unique for the registry -- and the handle for any object is a combination of the two. paf