Hi Jim,The goal shall remain to make non-breaking changes possible and breaking changes visible - with or without versioning extension.
I strongly believe most of the extensions will remain back-compatible, because it's healthy for the ecosystem. A coordinated version migration on scale of RDAP usage with multiple server instances and countless client instances not having any contract with each other is challenge comparable to DNS flag days. We don't want that happening for RDAP extensions.
Having told that, an extension identifier as a signal of an extension still being back compatible despite new version rolled out is a reasonable choice which will work in all cases. Versioning extension as an additional information what the version actually is with a facility to request a certain version, is a useful addition, but still optional addition from the protocol view.
draft-ietf-regext-rdap-extensions, as updating STD 95, is the only place where we can sharpen understanding of the contract between a client and the server as to what constitutes a breaking change and what not. Versioning extension draft cannot cover for it, unless it would also be an update to STD 95.
Kind Regards, Pawel On 11.06.26 18:05, Gould, James wrote:
Pawel,Thank you for defining what a breaking change is. The phrase “parse the response or misinterpret the response” is very broad and would assume not to include the addition of members. One issue is that many of the members are optional, so the lack of a member of an optional member should not be considered a breaking change. One aspect to consider is the viewpoint of a client that is expecting the new version of the extension, where if there is no signal (e.g., changing of the extension identifier for opaque versioning) of an interface change (e.g., adding new features and members) they may run into an issue with the broad “parse the response or misinterpret the response” language. I believe it’s best to change the opaque versioning for any interface change and not for just breaking change of the new extension from the old extension version. Adding new features is an interface change that warrants version signaling to me.I believe the approaches outlined in the sub-sections of 5.4 draft-ietf-regext-rdap-extensions apply only to opaque versioning and need to be moved into draft-ietf-regext-rdap-versioning, where appropriate. I don’t agree with the approaches outlined in section 5.4.3 and section 5.4.4. I believe it’s best for the server to support a default version of an extension and to support another version of the extension upon request, as defined in section 3.2 of draft-ietf-regext-rdap-versioning (query parameter or media type parameter). This approach has worked in EPP, where when the client supports multiple versions of an extension as defined by the login services, the server defaults to the latest version. This was done for the DNSSEC extension and the registry fee extension. Since RDAP doesn’t have the concept of login services, the server can default to the latest version of an extension and provide the information and the capability of the client to choose the version desired. We can add the relationships of the opaque extension versions into draft-ietf-regext-rdap-versioning to help with this for opaque versioning in the /help response.Thanks, -- JG cid87442*[email protected] *James Gould *Fellow Engineer[email protected] <applewebdata://13890C55-AAE8-4BF3-A6CE-B4BA42740803/[email protected]>703-948-3271 12061 Bluemont Way Reston, VA 20190 Verisign.com <http://verisigninc.com/> *From: *Pawel Kowalik <[email protected]> *Date: *Thursday, June 11, 2026 at 11:15 AM*To: *James Gould <[email protected]>, "[email protected]" <[email protected]>, "[email protected]" <[email protected]>, "[email protected]" <[email protected]> *Subject: *[EXTERNAL] Re: [regext] Re: draft-ietf-regext-rdap-x-media-type and draft-ietf-regext-rdap-versioning Compatibility Topic 1: Query IdentifiersHi Jim,a breaking change is the one that makes the client to either not being able to make a valid request, parse the response or misinterpret the response, which otherwisewould work with the previous version of the extension. It may be a change to format, syntax or semantics of an existing field,removing or renaming or a field (both mandatory and optional - the latter would let client believe a value is not there which may have a semantic meaning). There are similar considerations related to path segments, request parameters and referrals.What I notice now is that draft-ietf-regext-rdap-extensions-12 is in fact referring to a breaking change without really defining it. Only 5.4.2. is specifying what is a non-breaking change, which may imply that anything else is breaking. This may be fine, as long as 5.4.1 would tell it explicitly.Kind Regards, Pawel On 11.06.26 14:39, Gould, James wrote: Pawel, I want to understand what you're defining as a non-backward compatible change in RDAP, since most of the members returned are optional and the protocol requires ignoring unrecognized members, which differs from a provisioning protocol. Removing of a member should not break the client and renaming a member should similarly not break the client. Changing of the extension identifier could break the client if they need the extension using the old extension identifier. There is no way to foresee what will break a particular RDAP client. The best that the protocol can do is to provide support for the version information and to enable the client to discover and to indicate what extension version is desired. As pointed out previously, any RDAP extension that choses to use the point versioning type should include a normative reference to the versioning extension. A client and server adding support for a point versioning extension needs to also add support for the versioning extension. We can add language to the draft to reflect that requirement. A similar case would be made if an extension decided to use another versioning type defined by the versioning extension (e.g., extension identifier versioning that Scott brought up; although no official name has been given), with the exception of the opaque versioning that is there for backward compatibility and as an option for those that chose that type going forward. We may want to add relationships in the versioning extension for opaque version extensions to provide the visibility of changes, such as stating that "example1" is a successor of "example0" and that "example0" is a predecessor of "example1". Would this he helpful? In summary, the onboarding support of the versioning extension should be handled by the normative reference from new RDAP extensions that choose to use a type other than opaque versioning, so there should be no non-supporting clients of a point versioning RDAP extension. Thanks,--JG James Gould Fellow Engineer [email protected] <applewebdata://13890C55-AAE8-4BF3-A6CE-B4BA42740803/[email protected]> <mailto:applewebdata://13890C55-AAE8-4BF3-A6CE-B4BA42740803/[email protected]> 703-948-3271 12061 Bluemont Way Reston, VA 20190Verisign.com<http://verisigninc.com/> <http://secure-web.cisco.com/1Kjr6vhSbc6njskjsjxYUi8KOFBKvIdnK5vQS1s-v9N_MZgjrFSRd7H4cD5hSR3JwXrsUZH3uWbMODqllqDA_qQLPZyKaR_ZVfSQzOBflD8eyfzbXB24RJX_ydzGn2tZ8wgNERl7kArCNiBXSzczkggIK_gbQC_Fp6_VmOMkN4jBVPK5L6xfnuG6pNbUPpU2cRqYmqRKUIPS8mAwKP8IDzKPfxpV-sX-k1MxUrwo-zoqoH4piLBOW3vZ7CrkamXk08Gv58QJNOlJJ0ysyCyWt7v0rHYiaNoDeqkQCtHOBg-Y/http%3A%2F%2Fverisigninc.com%2F>On 6/11/26, 8:06 AM, "Pawel Kowalik" <[email protected] <mailto:[email protected]> <mailto:[email protected]>> wrote: Hi Jim, the question is about a scenario: - existing extension in version 1 in production, no versioning deployed because not yet specified (applied to any of existing extensions to date) - the extension upgrades to version 2, which is not back-compatible with 1, assuming extension identifier does not change as you propose and argument How would this scenario work for a) a client which implements versioning draft vs. client which does not b) a server which implements versioning draft vs. client which does not It's a simple matrix 2x2. You may add additional example of the same if version 2 has a hard requirement for implementers to also implement versioning. Kind Regards, Pawel On 11.06.26 13:45, Gould, James wrote: Andy, Inclusion of some query/response examples for supporting and non-supporting clients are reflected in the following draft-ietf-regext-rdap-versioning figures: 1. Non-supporting - Figure 8: RDAP Domain Lookup Response with No Client Specified Versioning 2. Supporting - Figure 9: RDAP Domain Lookup Response with "point_ext1-0.1" Extension Version 3. Supporting - Figure 10: RDAP Domain Lookup Response with "point_ext1-0.1" and "opaque_ext2" Extension Versions The only indication for the server whether the client is non-supporting or supporting is based on the inclusion of the Extension Versioning Request in Section 3.2. From an implementation perspective, a client will receive the "versioning_data" member that can be used by the supporting client to understand the extension versions included in the response and can be used for troubleshooting by a non-supporting client. The "versioning_data" member is in addition to the "rdapComformance" member and can be easily mapped based on indexing by the extension identifier. Since we have no experience with an extension transition in RDAP, we can look to the experience with an extension transition in EPP, which has been done many times. In EPP, our implementation was to return a default version based on the latest supported version of an extension. If a client does not support the latest supported version, based on the login services, the server returns the earlier supported version of the extension. For RDAP, the server can return the latest supported version of the extension and allow for the client to specify an earlier version of an extension using the Extension Versioning Request (query parameter or media type parameter) in Section 3.2 of draft-ietf-regext-rdap-versioning. The /help can also be used for automatic discovery of the extension versioning or manually for troubleshooting issues. The discovery can be done infrequently (e.g., daily), since the RDAP servers don't change their extension versions very often. I hope this helps. Thanks,
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ regext mailing list -- [email protected] To unsubscribe send an email to [email protected]
