> On Sep 5, 2023, at 16:25, Jeffrey Haas <jh...@pfrc.org> wrote: > > Acee, > > >> On Sep 5, 2023, at 12:42 PM, Acee Lindem <acee.i...@gmail.com> wrote: >> >> What I’d thought about for this requirement is an optional read-only leaf >> containing the hex representation of only the unknown bits. This would be >> simpler to model and would be fully backward compatible. >> >> For example: >> >> leaf unknown-flags { >> type yang:hex-string; >> description >> "When a bit is exchanged in the Graceful Restart Flags >> field that is unknown to this module, the hexadecimal >> representation of flags field is returned with the unknown >> bit(s)."; >> reference >> "RFC 4724: Graceful Restart Mechanism for BGP, Section >> 3."; >> } > > I'm not sure why this variant is particularly nicer than a set of bits in a > bit vector that themselves do not have any backward-incompatible changes. > > If I have, e.g. "unknown-3" and "unknown-4" as the options in the unknown > field, and then next rev comes along and defines 3 and the "known" bit gets > defined, deprecating unknown-3 is clear signaling that it's gone away. > > Part of the complaint in this thread is bit-3's contents "move" around. The > same would be true for the contents of a unknown field in a different non-bit > format: e.g. hex, uint, or binary.
The advantage is that the initial modeling is simpler (especially with lots of reserved bits) and an augmentation adding a definition of a bit that was previously defined doesn’t need to deprecate anything. Acee > > -- Jeff > _______________________________________________ netmod mailing list netmod@ietf.org https://www.ietf.org/mailman/listinfo/netmod