Re: [netmod] [Netconf] development plan

[Andy Bierman <[e-mail]>]

On Tue, Aug 6, 2013 at 3:55 AM, Ladislav Lhotka <[e-mail]> wrote:
> Juergen Schoenwaelder <[e-mail]> writes:
>
>> On Mon, Aug 05, 2013 at 10:04:38PM +0200, Martin Bjorklund wrote:
>>>
>>> FWIW, this is not how I understand I2RS - I think they explicitly
>>> mention the configuration data as being one thing, and that they want
>>> another "direct" method to modify the operational state.
>>>
>>
>> Yes, this is also my reading of draft-atlas-i2rs-architecture-01.txt,
>> which is likely going to become the I2RS architecture document soon.
>
> That document clearly states (Sec. 5.3): "... local device configuration is considered to be separate from the I2RS data store." So if Fig. 1 indicates that I2RS agents interact with Local Config, I assume it has to do so as a compliant NETCONF client. So "config true" data stay fully in the NETCONF realm.
>
> My conclusion from the discussion with the authors of draft-nitinb-i2rs-rib-info-model-01 is that I2RS data store is a subset of "config false" data from the NETCONF viewpoint, except that they are all read-only for NETCONF whereas the I2RS agent can write some of them.
>

I have not finished the code yet, so I will not comment on details,
but it should be quite possible to allow writable operational state
to be supported so it is compatible with NETCONF.

This meeting it was decided the injected state will not
be tied to local config (i.e., made to persist).  It only
lasts until explicitly removed or the next server reboot.

If you remember from the BoF, they want a fast path.
So implementing I2RS with RESTCONF or NETCONF is just
a matter of eliminating the slow parts ;-)

> Maybe I2RS folks reading this thread can shed more light on this issue.
>
> Lada
>
>>
>> /js
>>


Andy