On Fri, Sep 14, 2018 at 08:59:53AM +1200, Brian E Carpenter wrote: > Hi Ben, > > On 2018-09-14 01:34, Benjamin Kaduk wrote: > > On Wed, Sep 12, 2018 at 08:56:23AM +1200, Brian E Carpenter wrote: > >> On 2018-09-12 06:08, Robert Sparks wrote: > >>> I can live with this statement, but I don't like it. > >>> > >>> I'm in the camp that prefers the more specific "This changes the code > >>> you need to write" camp - I would prefer Update be restricted to the > >>> cases where you are changing the protocol defined in the updated > >>> document in an essential way. > >> > >> The IAB already opined indirectly on this topic. > >> https://tools.ietf.org/html/rfc6709#page-5 says: > >> > >>>>> Extension designers should examine their design for the following > >>>>> issues: > >>>>> > >>>>> 1. Modifications or extensions to the underlying protocol. An > >>>>> extension document should be considered to update the underlying > >>>>> protocol specification if an implementation of the underlying > >>>>> protocol would need to be updated to accommodate the extension. > >>>>> This should not be necessary if the underlying protocol was > >>>>> designed with a modular interface. > >> > >> (followed by more detailed discussion). > >> > >> I'm not sure the IESG text is 100% compatible with this. > > > > I believe it was intended to be, and I'm not sure what incompatibility you > > are seeing; could you say more? > > There's a subtlety in this phrase: > "would need to be updated to accommodate the extension." > Even though I'm a co-author of RFC6709, I'm not 100% sure what this means. > I think it means "if an implementor adds the extension to the code, she must > also modify the base code". The IAB text seems to be saying that if you add > a feature that does not *require* a change to the existing code, it's > not an update. I'm not sure I even agree with that, and I don't think > the IESG text does either. (If it does, we need a new tag "Extends:".) Ah, I see what you mean -- thanks. Our situation seems to be similar to the distinction between the converse and contrapositive in logic, though -- just because an Updates header is not necessary for this reason, does not a priori prevent an Updates header from being used for some other reason. -Ben
