On Thu, Apr 13, 2023 at 1:48 PM Jeffrey Haas <jh...@pfrc.org> wrote:
> Andy, > > On Apr 13, 2023, at 4:42 PM, Andy Bierman <a...@yumaworks.com> wrote: > >> >> Repeating my question to Acee... did you read the draft? This isn't a >> theoretical use case. >> > > Seeing no response, I'll assume "no". > > And yet, here you are stating an opinion. >> >> My opinion on this matter stems from the use case being mostly known and >> assigned bits and a small number of unknown bits and a desire to not to >> have to make my model users go fishing for the exceptions. >> > > > The typedef provides no semantics other than which bits are set in a bit > string. > > > For the unknown case. The semantic is implied by the leaf. > In all cases this is true. > > There are other ways to do that and all of them are valid usage of YANG. > I do not see how either encoding above requires "fishing" any more than > the other. > > > And thus ceteris paribus... you have no objections to this mechanism? > I do not support adoption or oppose adoption. Identification of arbitrary bits in a bit-string is a solved problem -- but not using YANG bits type. If people want that solution then they will be happy to have this new typedef. > >> And yet, we're here because the current design of YANG doesn't handle >> this unknown case well. >> >> > Changing the identifier breaks XML and JSON encoding of bits, so that is > why it says MUST NOT do this in RFC 7950. > > > The known leaf and unknown leaf do not change names. > The unknown leaf does not change type. > The known leaf's type is maintained according to current YANG versioning > rules. No on the wire behaviors are broken. > > Whether the semantic of a previously displayed unknown bit switching to a > known bit is a non-backward compatible change is a reasonable debate. I > suggest you offer an opinion in Jason's thread on that matter. > > The identifier is used on the wire. Changing it for a given leaf in a new module revision is not allowed. I guess I misunderstood that the bit identifier would change once it is known. e.g. bit-3 is changed to some other string. If the bit identifiers never change then there is no problem. > -- Jeff > Andy
_______________________________________________ netmod mailing list netmod@ietf.org https://www.ietf.org/mailman/listinfo/netmod
