Hi Pawel, Please find my comments inline, marked [JS2].
Thanks, Jasdip From: Pawel Kowalik <[email protected]> Date: Monday, December 15, 2025 at 4:08 AM To: Jasdip Singh <[email protected]>, Andy Newton <[email protected]>, [email protected] <[email protected]> Subject: Re: [regext] Re: I-D Action: draft-ietf-regext-rdap-extensions-09.txt On 11.12.25 22:02, Andy Newton wrote: Just to be clear, are you suggesting evolving an extension without an evolution signal should be allowed? In other words, not MUST NOT and not NOT RECOMMENDED? Yes, not MUST NOT and not NOT RECOMMENDED. But RECOMMEND using versioning extension and MUST assure clients do not break, whatever it means if not yet enough specified in STD 95. [JS] To your "assure clients do not break, whatever it means” point, would it help to clarify that in the draft? [PK] definitely. I am in doubt if we come up with an exhaustive list, but giving some guidance and examples would be of a value [JS2] Agreed. :) From a client’s perspective, changes injected into an existing extension (without a successor extension) could fall into: Breaking: new object class, referral with different semantics from another extension, altered JSON member value syntax, new required JSON member, new required query parameter [PK] "new object class" - only if returned on an already existing path, otherwise it's not breaking [JS2] OK. That seems reasonable given what Andy earlier said about an “objectClassName” member being the starting point for a client vis-a-vis an expected response. "referral with different semantics from another extension" - this I don't understand. What is "another extension"? Do you mean new semantics of an already defined referral? [JS2] Yes. More in line with the “MUST NOT” in section 5.2 (last paragraph). I believe otherwise the clients shall not break by a presence of any referral link, which they don't understand. [JS2] Agreed. "new required JSON member - this is typically valid for requests but not for responses, as "required" actually applies to the server not to the client. I would rather use "new JSON member which is critical to the client to properly understand the response”. [JS2] Yes, this is better. To further clarify, would “objectClassName” be considered such a critical member? Non-breaking: new JSON member, new query path, new query parameter, new HTTP header [PK] + New referral type [JS2] Right. By a new referral type, do you mean a new “rel” and “type” combination in a web link, or just the “type” in it?
_______________________________________________ regext mailing list -- [email protected] To unsubscribe send an email to [email protected]
