Hi, Jason
Thanks for kicking off some discussion around this question.
Please see my reply inline.

From: Sterne, Jason (Nokia - CA/Ottawa) [mailto:[email protected]]
Sent: Thursday, December 9, 2021 6:51 AM
To: maqiufang (A) <[email protected]>; [email protected]
Subject: RE: [netmod] Should the "with-origin" parameter be supported for 
<intended>?

Hi Qiufang,

In your first option, did you mean "understand a certain data node is from 
<running> or from <system>" ?
[Qiufang Ma] Yes, you're correct. We already define that a data node which is 
present in <system> and not overwritten by <running> will appear in 
<operational> with "origin=system" as always when applied successfully. So the 
server needs to remember whether a particular data node in <intended> is from 
<running> or from <system>. The question is should the server expose this 
source information to the client.

It is an interesting question about whether the origin annotation could/should 
be available in a read from <intended>, and what values that origin could take.

We should consider other transformations between <running> and <intended> 
(besides the merging of <system>):
- active/inactive config
- configuration templates

Inactive config would simply be stripped and not appear in <intended> at all. 
So nothing to discuss for that.

But would elements provided from within config templates have an 'origin' that 
indicates what template it came from ?
[Qiufang Ma] I am not quite sure what do you mean by "what template it came 
from". But for the client-defined template in <running>, elements provided from 
within the config template will exist in <intended> and finally have an 
origin="intended" when present in <operational>. I believe that this is already 
well defined in NMDA work.

I suppose the same question could apply to 'origin' in the <operational> DS.

Having origin purely available in the operational DS may not be complete 
enough. Some config nodes may not be present in operational because they are 
not "applied". So you wouldn't necessarily be able to get the full picture of 
the origin of all intended config by just checking the origin in the 
<operational> DS. Maybe that's an argument to have an origin in <intended> ?
[Qiufang Ma] That is a good point. <operational> only contains those are 
actually used by the system. It might be useful if a client can fetch 
<intended> can understand the origin of some data nodes which is after 
configuration transformation(e.g., client-defined/system-defined templates 
expansion) but may not be successfully applied by the device.

Best Regards,
Qiufang Ma

Jason

From: netmod <[email protected]<mailto:[email protected]>> On 
Behalf Of maqiufang (A)
Sent: Monday, November 22, 2021 7:11 AM
To: [email protected]<mailto:[email protected]>
Subject: [netmod] Should the "with-origin" parameter be supported for 
<intended>?

Hi, all

In the "system-defined configuration(draft-ma-netmod-with-system)" work, an 
optional datastore named "system" is introduced to hold non-deletable system 
configurations.
We define that if a server implements <intended>, <system> MUST be merged into 
<intended>.  So there is the cases that the clients can fetch <intended> and 
<intended> contains merged <system>.
The question is should we extend the "with-origin" parameter to support 
<intended>? The possible considerations for following two choices:
o  "with-origin" parameter should be supported for <intended>
?  It may be helpful for a client to fetch <intended> to understand a certain 
data node is from <running> or from <intended>
o  There is no need for <intended> to support "with-origin"
?  We already have <operational> to provide the origin for a particular data 
node
Any thoughts on this?


Best Regards,
Qiufang Ma

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

Reply via email to