Re: [Netconf] Solicit comments on inline action capability for NETCONF

[Robert Wilton <[e-mail]>]

Hi Qin,

Can you please provide some concrete example of the problem that needs 
to be solved here, and that can't be solved by existing tools.

I agree that transactional chained operations could be done, but I'm not 
convinced that they should be done.  E.g. if the problem can be solved 
in a simpler way then that it better.  E.g. considering enabling an 
interface and clearing counters at the same time.  Obviously if the 
interface is disabled, then a client could clear the counters first, 
then enable the interface, no transaction is required.

The more stuff that we add to NETCONF/YANG, the more complex it becomes 
and more expensive it becomes to implement and use.

So, for me, I think that a compelling problem statement would help 
evaluate whether this enhancement is really required.

Thanks,
Rob


On 19/07/2018 07:40, Qin Wu wrote:
> Hi, All:
> -----邮件原件-----
> 发件人: Robert Wilton [mailto:[e-mail]]
> 发送时间: 2018年7月17日 22:13
> 收件人: Robert Wilton <rwilton=[e-mail]>; Qin Wu <[e-mail]>; [e-mail]
> 主题: Re: [Netconf] Solicit comments on inline action capability for NETCONF
>
>
>
> On 17/07/2018 09:25, Juergen Schoenwaelder wrote:
>> On Tue, Jul 17, 2018 at 08:08:47AM -0400, Robert Wilton wrote:Hi Qi
>>> Hi Qin,
>>>
>>> Having read this draft, I can understand what the draft is proposing,
>>> but I don't currently understand why this is useful. Specifically, I
>>> don't find the example that is in the draft as compelling.  If the
>>> desire is to set the MTU and enable the interface as one
>>> configuration operation, then wouldn't the client just configure both
>>> mtu and enabled leaves at the same time.  Why is a separate action required here to enable the interface?
>>>
>> I have asked myself the same question multiple times. ;-)
>>
>> If people want engineer transactions consisting of multiple
>> operations, then they should do this for combinations of _arbitrary_
>> operations. At the end, edit-config is just an operation like
>> edit-data or any other operation.
> I agree.
>
> [Qin]: Good point, this is exactly what we propose to do. but in the current draft, we haven't made this clearly.
> In addition, we think action is still a special operation, need to be handled in the different way from other operations.
>
>> Personally, I do not think heavy weight transactions is the way to go
>> but if people want to engineer this, please make the solution at least
>> generic and not bound to edit-config or something like that.
> I also agree to both points.
>
> [Qin]: in our implementation practice,
> The action can be also be applied to get-config, get-data, so you are right, our intention
> Is to define it in the generic way.
> Thanks,
> Rob
>
>> /js
>>