To:
"'janusz@libertyrms.info'" <janusz@libertyrms.info>
Cc:
ietf-provreg@cafax.se, "Hollenbeck, Scott" <shollenbeck@verisign.com>
From:
"Gould, James" <JGould@verisign.com>
Date:
Tue, 2 Dec 2003 14:09:30 -0500
Sender:
owner-ietf-provreg@cafax.se
Subject:
RE: [ietf-provreg] RE: draft-hollenbeck-epp-rgp-01.txt comments/proposal
Janusz, When initially discussing this with Scott, my thought was that RGP could apply to more objects than just domains. I know of other provisioned objects that could benefit from an RGP like feature (i.e. E-mail Forwarding). In this, RGP would become another verb that that could be extended by command mappings similar to the core verbs (i.e. Domain RGP and E-mail Forwarding RGP) each with elements specific to each object. From a purest perspective this made sense, but I felt it was too complicated for now. By combining the object, in this case domain, with the protocol extension, I don't believe it invalidates the use of the protocol extension for adding a new verb. My intention of the initial e-mail was to propose the concept of the rgp protocol extension and not to define the final specification. Elements like <rgp:domain> and <rgp:period> could easily be added, since it would have its own XML schema. I disagree with your assertion that a protocol extension should not be directly associated with existing object mappings. Since you are defining a new verb, you can define exactly what objects it applies to. This could be an extensible object like described in the first paragraph or it could be a specific object like domain. Since domain is the predominate object for EPP, limiting the verbs to check, create, update, info, renew, delete, and transfer is too limiting. I hope some Registrars are following this so that we get their feedback. What we have here are two registries with different requirements for rgp (one with a renew feature and one without). If the specification is going to be registry independent it should support both cleanly. I don't like coupling rgp with update based on my previous comments related to confusion of what update elements should be accepted, but for me an update would match better than a renew. In our case, no updates are allowed during the redemption period. Any Registrar feedback? The three options right now include: RGP as renew command-response extension. The typical use case is: 1. Domain is auto renewed 2. Registrar deletes Domain via a <domain:delete> 3. Registrar sends a <domain:renew> with a <domain:period> of 0 (supporting a period is registry specific) with the rgp request extension 4. Registrar sends a <domain:renew> with a <domain:period> of 0 with the rgp report extension 5. Domain is ok RGP as update command-response extension. The typical use case is: 1. Domain is auto renewed 2. Registrar deletes Domain via a <domain:delete> 3. Registrar sends a <domain:update> with no updates (supporting update elements is registry specific) with the rgp request extension 4. Registrar sends a <domain:renew> with no updates (supporting update elements is registry specific) with the rgp report extension 5. Domain is ok RGP as protocol extension. The typical use case is: 1. Domain is auto renewed 2. Registrar deletes Domain via a <domain:delete> 3. Registrar sends an <rgp:request> with an optional <rgp:period> element. 4. Registrar sends an <rgp:report> 5. Domain is ok The command-response extension to the <domain:info> still applies with all three options. JG James F. Gould VeriSign Naming and Directory Services jgould@verisign.com -----Original Message----- From: janusz@libertyrms.info [mailto:janusz@libertyrms.info] Sent: Tuesday, December 02, 2003 12:26 PM To: JGould@verisign.com Cc: ietf-provreg@cafax.se; shollenbeck@verisign.com Subject: RE: [ietf-provreg] RE: draft-hollenbeck-epp-rgp-01.txt comments/proposal James, I don't think protocol extension offers more clarity than command response extension. In the first approach the verb is provided in explicit way but the object the verb is releted to is missing. Looking at the syntax you attached it is not obvious that <rgp:restore> is a domain related operation. In the second approach the object is clearly defined but the verb is defined in implicit manner. The protocol extension approach is more difficult to implement. With this approach the rgp syntax will be more complex. The samples you attached to your original proposal are incorrect and they offer very simplistic view of necessary changes to rgp draft document. I pointed that out in my first response. More complex rgp syntax can be used as a rough but objective measure for determining amount of efforts required for implementing a particular approach. Subjective references to a particular server implementation attempts SHOULD NOT be used as a such measure. You said in one of your responses in the thread that if protocol extension is not adopted for rgp then "there is really no purpose for defining the protocol extension in EPP". I disagree with that judgment. An example for a usefull application of protocol extension would a be a new verb that is not directly associated with existing object mappings. Something like poll or hello command. Janusz Sienkiewicz