[as a contributor]

Hi Rob,

>I don¹t agree with this diagram. The applied configuration is essentially
>the current running state of the hardware. It cannot be directly changed
>by a user, and hence it needs to be read-only and hence config false.

Agreed that the applied config is read-only, but that doesn't mean the
data tree has to be config false.  For instance, in the draft submitted
yesterday, this requirement is achieved by defining a read-only datastore.


>operational: true is used to represent op-state that does not correspond
>to any configuration (neither applied, nor intended), but represent
>parameters that are related to protocol interactions, counters, etc. This
>again is read-only.

I'm unclear with the difference, how is it different than data modeled
config false?


>>Consider a system that has a single NMS that talks to all elements, and
>>it is the *only* means of managing those elements. In general, that NMS
>>knows what the state of the intended config is (it wrote it), if the
>>system is synchronous then it also knows the state of the applied
>>config. Therefore if it is polling, it is likely to want to only get the
>>derived state data, which can be filtered on operational true. The
>>intention of these statements is to impress the need on the reader to
>>have a mechanism to poll only this data efficiently (if this is, indeed,
>>to be polled).

Okay, but there are other ways to enable a client to poll just the derived
state (config false nodes).  For instance, the draft posted yesterday
introduces a <get-state> operation to do this.


K.

_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod

Reply via email to