Re: [netmod] evaluation of "when" under NMDA

[Robert Wilton <[e-mail]>]

On 05/12/2017 14:38, Martin Bjorklund wrote:
> Balazs Lengyel <[e-mail]> wrote:
>> On 2017-12-05 11:04, Robert Wilton wrote:
>>
>>
>>
>>
>>
>>          I see your point now.
>>          The server has to evaluate the when-stmts in operational.
>>
>>      I think that this is probably down to implementation, but I don't
>>      think that this is necessarily required. A server is meant to
>>      conform to 'when' statements in <operational> (e.g. if the system
>>      is in a normal steady state), but they are allowed to be violated,
>>      and I'm not expecting that a server would evaluate them (except
>>      perhaps to discover implementation bugs). Further, if violations
>>      of when statements in <operational> are detected then I don't
>>      think that there is anything that the server can reasonable do.
>>
>>
>>
>> BALAZS: I always thought that if a when statement's argument was true
>> but becomes false, all instance data that is set/written according to
>> the schema nodes affected by the when statement shall be removed by
>> the server. So IMHO the server can and should do something about a
>> violated when statement.
> Yes.
But I still don't think that requires that you have to evaluate when 
constraints on <operational>.

E.g. [implementation dependent]:
  - user sends config to <running>
  - server calculates change to <intended> including implicit deletes 
due to now false when statements.
  - server sends calculated config change to daemons.
  - config daemons late update operational with the updated applied 
configuration.  When statements can be transiently violated in 
operational, but should converge to the behavior defined in the data 
model, all being well.

Thanks,
Rob


>
>> Actually I would like a list of statements and constraints that MUST
>> be satisfied in the different data stores. Speaking about syntactic
>> versus semantic seems fluffy.
> See RFC 7950, section 8 (esp. 8.1 and 8.2).
>
>
> /martin
> .
>