From: Martin J. Dürst <duerst@xxxxxxxxxxxxxxx> Sent: 26 November 2023 23:28 Hello Tom, others, On 2023-11-26 20:15, tom petch wrote: <tp> > A further example that I omitted in my first post, is about XML. RFC with YANG modules are encouraged to include examples and that has been XML. Sometime after that started, I started seeing posts from IANA to WG lists asking an XML group to review the examples and the XML group always said that the examples were invalid and wanted changes. As far as YANG is concerned, the examples are perfectly valid and pass validation by tools but authors started to incorporate those changes to placate IANA. > > What is IANA's justification for requiring this conformance to someone else's, not YANG's, rules? 'valid'/'invalid' for XML have very clear definitions. If you use XML, you should abide by XML's rules, I'd assume. It's possible to not use validation (when not using a DTD (or a schema)). I don't know whether YANG uses DTDs or schemas. If it does, it should abide by its own rules. If it doesn't, you are talking about something else and should use different terminology and explain the problems in greater detail. <tp2> I think that the root cause is that the designers of what became YANG did not have the in-depth knowledge of XML that others have. A related problem to the one I am raising here is that many RFC are invalid in the sense of producing the wrong answer because the scope of XPath varies between absolute and relative form and the relative form is such a pain to get right when it is nested ten deep that YANG module authors used the absolute form not realising that that would give positive results with XPath which were not wanted ie a test for the presence of the OSPF protocol on a specific link would yield a positive whenever any link in the device had the OSPF protocol. The result is valid YANG and a device that works but looks a mess when trying to see what is going on because every link has been augmented with OSPF parameters instead of just the OSPF ones. This surfaced on the Netmod list when someone with a better knowledge of XML particpated and is now a well-known problem. Back to the case I raised. Look at NETCONF mail list [netconf] [IANA #1261213] expert review for draft-ietf-netconf-https-notif (ns) Tim Bray 2022-11-18 which is one of several such responses to various requests by IANA. Tim says 'This is up to the usual standard for YANG documents.' I am unclear whether this is sarcasm but I do understand the I-D author's response of 'As for making the requested change, it does not make sense to me (as author) to introduce such language now, given the history'. My paraphrase of the situation is why have IANA started intervening. The history is that we were getting on fine, without the changes requested by Tim, until they did! To me this is similar but different to Carsten's comments about the problems on the OPSAWG list which arise from not knowing what IANA can and can not do (e.g. W.r.t. ISE documents). HTH Tom Petch > What I now see is a lack of XML for examples and excessive use of JSON. (Will IANA find a JSON directorate to declare these invalid:-( > > For me, XML is superior to JSON when it comes to illustrating the use of YANG and I am thinking that if IANA had not intervened, then I would still be getting XML. I may be wrong, but I don't think IANA has any preferences for XML or JSON. YANG itself defines whether it can be written in both XML and JSON, or in only one of them. If it allows JSON, then there shouldn't be any problem with JSON, even if you personally prefer XML. Regards, Martin.
