From: netmod <[email protected]> on behalf of Jensen Zhang <[email protected]> Sent: 28 June 2022 01:38
Hi all, We are working on an I-D [1] that requires enumeration typedefs for some IANA registries. <tp> I do not understand that statement. An IANA registry is a namespace with values, numeric or sometimes character string, listing possible values and giving references for the definition of the values and the rules for the maintenance of that namespace. Some WG put their namespaces into IANA, others do not - if only one WG is involved, especially if there is only one base document involved, then arguably there are costs and no value in involving IANA. If a registry is set up with IANA, then I cannot see a registry requiring 'enumeration typedefs', enumeration is a data modelling concept, a list of possible values with names (simplifying). I cannot see a registry requiring enumerations. Historically, the IETF used SMI which is mostly numeric but with type of enumeration there was an associated text string. YANG has a type enumeration which uses text string and any associated number is documentation only. Some users of YANG do not seem to understand this and think that name and number are bound together - they are not. This causes problems when the protocol is numeric, as most are, You then need an authoritative mapping between name and number which YANG does not provide. Having IANA maintain a YANG module requires IANA to have the YANG skills to perform this. On more than one occasion recently, they seem to have lacked those skills (good as they are at maintaining registries in the face of unclear, confused, contradictory information in RFC). The registry policy is a significant factor - if that is RFC required, then there seems little point asking IANA to do anything. It is only with such as Expert Review that IANA's involvement may be beneficial, but the experts need the expertise, not just in the protocol but in YANG. Finally, some WG choose identity, others enumeration in their modelling. One key difference is that with YANG identity, you in a sense give up change control - anyone else can do anything - so a registry of YANG identity seems like an oxymoron. I see your statement as an incomplete solution to some problem. What is the problem? Tom Petch We notice that some modules [2] self-contain identities and typedefs for IANA registries, but some other modules [3] create separated `iana-xxx-types.yang` modules. We wonder what is the best practice on when to create IANA-maintained YANG modules. Could anyone give us any guidance? [1] https://github.com/ietf-wg-alto/draft-ietf-alto-oam-yang/issues/1 [2] https://github.com/YangModels/yang/blob/main/experimental/ietf-extracted-YANG-modules/ietf-ospf-srv6%402018-10-11.yang [3] https://github.com/YangModels/yang/blob/main/standard/ietf/RFC/iana-routing-types%402018-10-29.yang Thanks, Jensen _______________________________________________ netmod mailing list [email protected] https://www.ietf.org/mailman/listinfo/netmod
