On 3/26/19 03:22, Juergen Schoenwaelder wrote:
> On Tue, Mar 26, 2019 at 03:12:26AM -0400, Joe Clarke wrote:
>> On 3/26/19 01:51, Juergen Schoenwaelder wrote:
>>> Qin,
>>>
>>> the idea should be to make things simpler, not more complex. Perhaps
>>> it is not necessary to expose N options to reset a device. Perhaps a
>>> simple "factory-reset" RPC which resets all relevant datastores in an
>>> implementation specific manner is sufficient. Why expose more details
>>> to the management client?
>>
>> This would certainly make it simpler from the RPC standpoint.  However,
>> if one can <get-data> from the factory-default DSes, I still think there
>> is a need to know what factory-default DS maps to what other DS (in the
>> case where there might be multiple that are different).
>>
> 
> The notion of multiple factory-default datastores sounds complex. And
> what is a management application going to do with them? How would a
> management application know which sets of datastores to reset together
> in a meaningful way?

This is why I think having a single RPC to "reset the device to factory"
makes sense.  Not to say that's the only use case, but it would be
useful in a number of cases.

> 
> My naive interpretation of the factory default DS (a single one) would
> be that it exposes the content you will find in <running> after the
> factory-reset has been executed. An extended version of Figure 2 of
> RFC 8342 would look like this:

Yep, this makes sense.  And I fully admit that my use case is likely
vendor-specific, but we may not be the only vendor with ancillary
datastores that do not look like <running> or <startup>.

Joe

> 
>      +-------------+                 +-----------+        +-----------+
>      | <candidate> |                 | <startup> |        | <factory> |
>      |  (ct, rw)   |<---+       +--->| (ct, rw)  |        | (ct, ro)  |
>      +-------------+    |       |    +-----------+        +-----------+
>             |           |       |           |                   |
>             |         +-----------+         |                   |
>             +-------->| <running> |<--------+                   |
>                       | (ct, rw)  |<----------------------------+
>                       +-----------+
> 
> And exposing such a factory default datastore would be optional I
> think.
> 
> /js
> 

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

Reply via email to