Re: [yang-doctors] [netmod] Operational State usage of YANG choices and constraints

Hi Andy, 

> On Jan 19, 2024, at 12:17, Andy Bierman <andy@yumaworks.com> wrote:
> 
> 
> 
> On Fri, Jan 19, 2024 at 9:02 AM Acee Lindem <acee.ietf@gmail.com <mailto:acee.ietf@gmail.com>> wrote:
>> Along the same lines, what is the sentiment for “mandatory true” data nodes for operational state (i.e., “config false” nodes)? Does this make sense to identify data nodes the must be returned if the parent node is returned?
>> 
> 
> 
> What does "must be returned" mean?
> 
> Obviously, it does not mean returning the data even if the filters do not select that node. (I hope)
> I used to think it meant conformance  (whether the server must implement or may implement).
> I was told years ago that is not the case. Use if-feature for that. No if-feature == mandatory-to-implement.
> 
> If a client requests the parent subtree, then how would it know the difference between config=false
> nodes that are not present vs. the server just decided not to return them (even though conformance
> to the retrieval operation requires them)?
> 
> IMO the mandatory-stmt for config=false nodes does not make any sense at all.

I’m fine with that - we just need to make sure that this is reflected in our IETF models. I think these points are worth addressing in RFC8407 BIS. 

Thanks,
Acee

> 
>  
>> Thanks
>> Acee
>> 
> 
> Andy
>  
>> > On Dec 22, 2023, at 2:36 PM, Jürgen Schönwälder <jschoenwaelder@constructor.university> wrote:
>> > 
>> > On Fri, Dec 22, 2023 at 07:22:55PM +0000, Kent Watsen wrote:
>> >> With limited experience wrt the impact on servers, as a client, it’s always best for the opstate data to be modeled as accurately as possible, for better processing and user experience.
>> >> 
>> > 
>> > What is accurate?
>> > 
>> > I think the answer is "it depends". There are states that a model
>> > allows to represent and there are states it does not allow to
>> > represent. If a device ends up in a state that the model can't
>> > represent, then the device has a problem, From a debugging point of
>> > view, the worst is a device in a state that can't be represented
>> > propoerly reporting a valid state it is not in.
>> > 
>> > So like everything else, it is a modeling decision, like picking types
>> > and everything else. I am not sure that 'as accurate as possible" is a
>> > helpful guideline; for operational state I prefer to see as much as
>> > possible the device's true state. (But even picking data types for
>> > leaves restricts what can be represented, so it is a judgement call.)
>> > 
>> > /js
>> > 
>> > -- 
>> > Jürgen Schönwälder              Constructor University Bremen gGmbH
>> > Phone: +49 421 200 3587         Campus Ring 1 | 28759 Bremen | Germany
>> > Fax:   +49 421 200 3103         <https://constructor.university/>
>> 
>> _______________________________________________
>> yang-doctors mailing list
>> yang-doctors@ietf.org <mailto:yang-doctors@ietf.org>
>> https://www.ietf.org/mailman/listinfo/yang-doctors