Pawel, RFC 9082 states the following for the Help Path Segment:
The help path segment can be used to request helpful information (command syntax, terms of service, privacy policy, rate-limiting policy, supported authentication methods, supported extensions, technical support contact, etc.) from an RDAP server. The response to "help" should provide basic information that a client needs to successfully use the service. This is exactly what the versioning extension is doing in the "versioning_help" member, by providing help information on supported extensions. A client is not required to use the versioning extension to perform queries, since the server does have a default extension version that is specified in the "versioning_help" member. If a client does run into an extension compatibility issue, it could use the help command to programmically (lazily) or manually determine the root cause of the issue for resolution. There is no requirement or expectation that an RDAP client implement 2-RTT. Thanks, -- JG James Gould Fellow Engineer jgo...@verisign.com <applewebdata://13890C55-AAE8-4BF3-A6CE-B4BA42740803/jgo...@verisign.com> 703-948-3271 12061 Bluemont Way Reston, VA 20190 Verisign.com <http://verisigninc.com/> On 11/21/24, 12:19 PM, "Pawel Kowalik" <kowa...@denic.de <mailto:kowa...@denic.de>> wrote: Hi, Thinking of interactions between the client and server that versioning draft assumes I think we are heading towards 2-RTT model for every request. Step 1: The client makes an HTTP GET request to the /help endpoint of the RDAP server. Step 2: The response is processed to extract rdapConformance and versioning. Step 3: Compare the server's supported extensions and versions with those supported by the client. Step 4: If compatible configurations are found, the client makes target request to a resource endpoint (e.g., domain/foo.com), using headers or query parameters to specify the desired configuration. So we have 2 full RTTs. Of course a client can cache it for some time but not forever, as the server may change at any time its configuration. In a cold state or a client without capability to cache this will be always 2-RTT if the client would like to be aware of the versions. I don't think this is in line with rfc7480, which assumes "a client implementation should be possible using common operating system scripting tools (e.g., bash and wget)". Also not with the usage pattern defined in Section 1. None of it is normative, however I would not just ignore it without discussing consequences of going this way. I would like to see more that versioning draft would assume 1-RTT model as a primary use-case, so that the client would put a range of versions it supports into a request and the server would put best efforts to fulfil it. This would be also more "redirect friendly", so that a redirected request to another server (no matter if with query parameters or headers) would have the same information about client capabilities and able to serve the response as opposed to getting a request for configuration crafted for the origin server of a redirect. Kind Regards, Pawel _______________________________________________ regext mailing list -- regext@ietf.org To unsubscribe send an email to regext-le...@ietf.org
