Hi Mario, On Wed, Jul 24, 2019 at 08:08:08PM +0200, Mario Loffredo wrote: > Il 24/07/2019 16:29, Tom Harrison ha scritto: >> - Section 5 of RFC 7483 suggests that objects should always include a >> 'self' link, regardless of whether they are top-level objects, >> embedded objects, or search result objects. I think that including >> 'self' links for all objects, even those returned in response to an >> 'id' fieldset search, would be useful, because it saves the client >> from having to re-search when they want further details for only >> a small number of objects in a set of search results. > > Your opinion disagrees with what is written in > draft-blanchet-regext-rdap-deployfindings-05. I, too, agree with > Marc that self links can cause inifinite loops when the user is a > software agent
I think that document's content is about links that refer to the current object, but that appear with a link relation type other than 'self'. If the link has a link relation type of 'self', then the client should skip it during link processing. > but, at the same time, I recognize that, when the id field set is > applied, a self link for each result would be helpful to a human. > > Definitively, I think that such a feature could be provided by an > RDAP client smarter than a simple web browser. This part of my response is a bit awkward, because it spans multiple documents. I forgot to include in my reverse-search mail that it would be good to expand the supported object types to include IP networks and ASNs, since those are the main searches that we (APNIC) are looking to support. Assuming that's OK, then including 'self' links in partial responses is more useful than for domain/nameserver/entity alone, because for IP networks it will save the client from having to calculate the CIDR range based on the startAddress and endAddress values (querying by the exact CIDR range is necessary because of hierarchical delegations). Given that including 'self' links will increase response sizes substantially, and that the need for 'self' links will not be consistent across clients or object types, I think that your suggestion from the other mail about having an 'extendedId' fieldset that includes 'self' links sounds like a good way of dealing with this issue. -Tom _______________________________________________ regext mailing list regext@ietf.org https://www.ietf.org/mailman/listinfo/regext
